Alarm data analysis method and system of power plant

Abstract

The invention relates to the technical field of data analysis, in particular to an alarm data analysis method and system of a power plant. The method comprises the following steps: collecting original alarm signals, generating a characteristic matrix of the alarm signals through frequency calculation and time sequence characteristic analysis, carrying out visualization processing, positioning repeated alarm points, carrying out clustering analysis, generating classified alarm data, carrying out inter-class correlation analysis, and generating deduced alarm optimization data. Performing pattern recognition on the feature matrix based on an alarm historical record, extracting an alarm pattern sequence, predicting an alarm flooding situation, performing response evaluation on the classified alarm data, optimizing and adjusting the deduction data in combination with a flooding situation prediction result to generate pre-adjusted alarm optimization data, and outputting the pre-adjusted alarm optimization data. And integrating and summarizing the classification data, the response evaluation data and the optimization data to form an alarm analysis report. According to the invention, a safer and more efficient alarm data analysis method of the power plant is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of data analysis, and in particular to an alarm data analysis method and system for a power plant. Background Art

[0002] Alarm data analysis is an important part of ensuring the safe and stable operation of the power system. However, existing alarm management methods often focus on the recording and static analysis of single-point data, lacking a comprehensive exploration of the frequency, timing characteristics and correlation of alarm data, making it difficult to accurately identify repeated alarm points and potential systemic problems. In addition, due to the complexity and frequency of alarm types, the traditional manual analysis model shows obvious efficiency bottlenecks and insufficient accuracy when facing high-frequency, multi-type alarm signals. Most of the current research focuses on the analysis of a single alarm dimension, while the comprehensive evaluation of multi-dimensional data is still less explored. Especially in the scenario of dealing with repeated alarms and high-frequency alarms, it is impossible to effectively classify and model the continuity, category and correlation of alarms, which limits the implementation effect of alarm optimization. At the same time, there is a lack of reliable timing modeling and dynamic adjustment strategies for the identification of alarm patterns and the prediction of flooding trends, which makes the prediction accuracy of alarm data low. Summary of the invention

[0003] Based on this, it is necessary to provide an alarm data analysis method and system for a power plant to solve at least one of the above technical problems.

[0004] To achieve the above object, a method for analyzing alarm data of a power plant comprises the following steps:

[0005] Step S1: collecting the original alarm signal; performing frequency calculation on the original alarm signal to obtain the frequency characteristics of the alarm signal; performing time series characteristic analysis on the frequency characteristics of the alarm signal to generate a time series alarm characteristic matrix, and performing visualization processing;

[0006] Step S2: locating repeated alarm points on the time series alarm feature matrix to obtain repeated alarm points; clustering the alarm types of the repeated alarm points based on a preset evaluation time threshold to generate classified alarm data;

[0007] Step S3: Based on the classified alarm data, inter-class correlation analysis is performed on the repeated alarm points to obtain alarm point correlation data; based on the alarm point correlation data, alarm optimization deduction is performed to generate deduced alarm optimization data;

[0008] Step S4: Acquire alarm history records; perform alarm pattern recognition on the time series alarm feature matrix according to the alarm history records to obtain an alarm pattern sequence; perform flooding situation prediction on the alarm pattern sequence to generate a predicted alarm flooding situation;

[0009] Step S5: performing alarm response evaluation on the classified alarm data based on the alarm history records to generate response evaluation data; pre-adjusting the deduced alarm optimization data based on the predicted alarm flooding situation and the response evaluation data to obtain pre-adjusted alarm optimization data;

[0010] Step S6: Summarize the classified alarm data, response evaluation data and pre-adjustment alarm optimization data to generate an alarm analysis report.

[0011] The present invention can extract the key features of the alarm signal by frequency calculation and time series feature analysis of the original alarm signal. The generated time series alarm feature matrix provides an intuitive data basis for subsequent analysis. The positioning of repeated alarm points and clustering of alarm types realize the effective classification of alarm events. The generated classified alarm data provides support for subsequent correlation analysis. The implementation of inter-class correlation analysis can identify the relationship between alarm points. The obtained alarm point correlation data provides a basis for optimization deduction. The implementation of alarm optimization deduction can generate more accurate optimization strategies. The combination of obtaining alarm history records and alarm pattern recognition can reveal long-term trends. The generated alarm pattern sequence provides a data basis for flooding situation prediction. The generation of response evaluation data can evaluate the effect of alarm response. Alarm response evaluation is performed for optimizing alarm data. Response evaluation data is generated, and the deduction data is adjusted in combination with the prediction results to generate pre-adjustment optimization data. The alarm optimization decision is dynamically adjusted. The classified alarm data, response evaluation data and pre-adjustment optimization data are summarized and analyzed to generate an alarm analysis report, which provides systematic comprehensive evaluation data and guidance basis for equipment alarm management and optimization.

[0012] Preferably, step S1 comprises the following steps:

[0013] Step S11: collecting the original alarm signal; performing digital filtering on the original alarm signal to obtain a filtered alarm signal;

[0014] Step S12: segmenting the filtered alarm signal into time windows to obtain time period alarm data; performing overlapping window processing on the time period alarm data to obtain overlapping time period alarm data;

[0015] Step S13: Perform frequency statistics processing on the alarm data of the overlapping period to generate the frequency characteristics of the alarm signal; perform time series correlation calculation on the frequency characteristics of the alarm signal to obtain time series characteristic data;

[0016] Step S14: performing matrix conversion processing on the time series feature data to obtain a time series alarm feature matrix; and visually displaying the time series alarm feature matrix.

[0017] The present invention effectively removes signal noise and enhances signal effectiveness through the collection and digital filtering of original alarm signals. It improves the coverage and continuity of time period alarm data by performing time window segmentation and overlapping window processing on filtered alarm signals. It extracts frequency characteristics and time series characteristic data of alarm signals by performing frequency statistics and time series correlation calculation on alarm data of overlapping time periods, thereby improving the feature recognition accuracy of alarm signals. It enhances the structured expression and visual analysis capabilities of alarm characteristics and improves the comprehensiveness and analysis efficiency of alarm data by performing matrix conversion and visual display on time series characteristic data.

[0018] Preferably, step S2 comprises the following steps:

[0019] Step S21: Perform frequency statistical analysis on the time series alarm feature matrix to obtain alarm frequency data;

[0020] Step S22: performing threshold comparison and screening on the alarm frequency data according to a preset alarm frequency threshold to obtain repeated alarm data; locating repeated alarm points according to the repeated alarm data to obtain repeated alarm points;

[0021] Step S23: performing time dimension aggregation on repeated alarm points to generate time clustering data; performing feature focusing on the time clustering data to obtain feature clustering data;

[0022] Step S24: classify the feature clustering data by type labeling to generate classified alarm data.

[0023] The present invention effectively captures the frequency characteristics of alarm signals and improves the accuracy of data analysis through frequency statistical analysis of the time series alarm feature matrix. It compares and screens the alarm frequency data by pre-setting the alarm frequency threshold, realizes the precise extraction of repeated alarm data, and enhances the ability to locate repeated alarm points. It generates time clustering data through time dimension aggregation, strengthens the time correlation and concentration of alarm data, obtains feature clustering data through feature focusing processing, improves the extraction depth and analysis pertinence of alarm feature data, generates classified alarm data through type tag classification, and optimizes the classification structure and management efficiency of alarm data.

[0024] Preferably, step S3 comprises the following steps:

[0025] Step S31: constructing a feature vector for the classified alarm data to obtain feature description data; performing similarity calculation on the feature description data to generate a similarity matrix;

[0026] Step S32: performing inter-class correlation analysis on repeated alarm points based on the similarity matrix to obtain alarm point correlation data;

[0027] Step S33: performing optimization rule matching on the alarm point correlation data to obtain optimization strategy data; performing numerical simulation test based on the optimization strategy data to generate numerical simulation results;

[0028] Step S34: Perform alarm optimization deduction on the optimization strategy data based on the numerical simulation results to generate deduced alarm optimization data.

[0029] The present invention improves the feature description capability of alarm data by constructing feature vectors for classified alarm data, generates a similarity matrix through similarity calculation, accurately quantifies the similarity relationship between alarm data, performs inter-class correlation analysis on repeated alarm points based on the similarity matrix, reveals the potential correlation between alarm points, optimizes alarm point correlation data using optimization rule matching, generates optimization strategy data, verifies the optimization strategy data through numerical simulation tests, enhances the scientific nature and operability of the optimization strategy, performs alarm optimization deduction on the optimization strategy data based on the numerical simulation results, and improves the accuracy and practicality of alarm management.

[0030] Preferably, step S32 includes the following steps:

[0031] Based on the similarity matrix, the repeated alarm points are associated and mapped to obtain the alarm point dependency matrix; based on the alarm point dependency matrix, the repeated alarm points are associated and analyzed to generate the alarm point associated trajectory;

[0032] Perform aggregation analysis based on alarm point correlation trajectories to obtain alarm point behavior patterns; extract inter-point behavior rules from alarm point behavior patterns to obtain inter-point behavior rules;

[0033] The separation degree of repeated alarm points is calculated based on the behavior rules between points to generate the inter-class sparsity;

[0034] A multi-layer correlation network is constructed according to the inter-class sparsity and alarm point dependency matrix to obtain the alarm point association network; the alarm point association network is analyzed layer by layer to generate alarm point correlation data.

[0035] The present invention performs association mapping on repeated alarm points based on a similarity matrix to obtain an alarm point dependency matrix, accurately reveals the mutual dependence between alarm points, generates alarm point association trajectories by performing association trajectory analysis according to the alarm point dependency matrix, comprehensively reflects the temporal correlation of alarm points, obtains alarm point behavior patterns by performing aggregation analysis on alarm point association trajectories, clarifies the potential behavior laws of alarm points, obtains inter-point behavior laws by extracting inter-point behavior laws, reveals the behavioral similarities and differences between alarm points, calculates the degree of separation of repeated alarm points based on inter-point behavior laws, generates inter-class sparsity, reflects the degree of separation between different alarm points, constructs a multi-layer correlation network according to inter-class sparsity and the alarm point dependency matrix, obtains an alarm point association network, comprehensively analyzes the multi-level relationships between alarm points, generates alarm point correlation data by parsing the alarm point association network layer by layer, and optimizes the correlation analysis and management of alarm points.

[0036] Preferably, step S4 comprises the following steps:

[0037] Step S41: Acquire alarm history records; perform time completion processing on the alarm history records to obtain completed history records;

[0038] Step S42: performing periodic analysis on the time series alarm feature matrix based on the completed historical records to generate periodic feature data; performing alarm trend recognition on the periodic feature data to obtain trend feature data;

[0039] Step S43: performing pattern decomposition on the trend feature data to obtain pattern component data; performing pattern matching on the pattern component data and the time series alarm feature matrix to generate an alarm pattern sequence;

[0040] Step S44: Decompose the alarm pattern sequence to obtain pattern sequence features; perform time series change trend analysis on the pattern sequence features to generate alarm pattern change trends;

[0041] Step S45: Perform situation assessment based on the alarm mode change trend to generate a predicted alarm flooding situation.

[0042] The present invention performs time completion processing on alarm historical records to obtain completed historical records, thereby ensuring the integrity and continuity of data; performs periodic analysis on a time series alarm feature matrix based on the completed historical records to generate periodic feature data, thereby revealing the periodic characteristics of the alarm data; performs alarm trend identification on the periodic feature data to obtain trend feature data, thereby analyzing the long-term change trend of the alarm signal; performs pattern decomposition on the trend feature data to obtain pattern component data, thereby accurately extracting potential pattern components in the alarm data; performs pattern matching on the pattern component data and the time series alarm feature matrix to generate an alarm pattern sequence, thereby ensuring effective identification and classification of the alarm pattern; performs feature decomposition on the alarm pattern sequence to obtain pattern sequence features, thereby extracting the key features of the alarm pattern in detail; performs time series change trend analysis on the pattern sequence features to generate an alarm pattern change trend, thereby analyzing the change law of the alarm pattern over time; performs situation assessment based on the alarm pattern change trend, thereby generating a predicted alarm flooding situation.

[0043] Preferably, step S5 comprises the following steps:

[0044] Step S51: Analyze the response time of the alarm history records to obtain historical response time data;

[0045] Step S52: mapping the classified alarm data to a response time according to the historical response time data to generate a predicted current alarm response time;

[0046] Step S53: Use the preset index and alarm scoring formula to perform alarm response evaluation on the predicted current alarm response time to generate response evaluation data, wherein the alarm scoring formula is:

[0047]

[0048] Among them, S is the final score, V is the achieved value of the device, and T is the index of this value;

[0049] Step S54: self-organizing mapping is performed on the predicted alarm flooding situation and response evaluation data to obtain mapping relationship data; and pre-adjusting the deduced alarm optimization data based on the mapping relationship data to obtain pre-adjusted alarm optimization data.

[0050] The present invention can accurately capture the trend and distribution of historical alarm response time by analyzing the response time of alarm history records, provide data support for the prediction of current alarm response time, and can predict the response time of current alarm based on historical response time data by mapping the response time of classified alarm data, thereby enhancing the accuracy and timeliness of alarm response evaluation. By using preset indicators and alarm scoring formulas, the quantitative evaluation of current alarm response time can be achieved, and then the scoring formula is used: The performance of the alarm response is quantified and scored. In the formula, S represents the final score, V represents the achieved value of the device, and T represents the indicator of the value. The formula calculation can effectively avoid over-response and non-response. Combined with the scoring results, it is helpful to accurately judge the efficiency and reliability of the alarm response. By performing self-organizing mapping on the predicted alarm flooding situation and response evaluation data, a more reasonable alarm optimization plan can be obtained based on historical data and real-time response evaluation. After pre-adjustment, the optimized alarm data can better reflect the actual alarm processing needs of the power plant, ensure the timeliness and accuracy of the alarm data, and ultimately improve the efficiency of the power plant alarm management, reduce human processing deviations, and optimize the overall performance of equipment management and fault response.

[0051] Preferably, step S54 includes the following steps:

[0052] Adaptively segment the predicted alarm flooding situation by thresholds to obtain flooding level data; decompose the response evaluation data into time segments to obtain time response characteristics;

[0053] Cross-validate the flooding level data and time response characteristics to obtain the response weight matrix;

[0054] Perform multi-dimensional decomposition on the deduction alarm optimization data to obtain the optimized feature vector; perform self-organizing mapping on the optimized feature vector according to the response weight matrix to obtain mapping relationship data;

[0055] Compensation condition analysis is performed based on flooding level data and time response characteristics to generate compensation condition parameters; compensation characteristic conversion is performed on the compensation condition parameters based on mapping relationship data to generate compensation characteristic vectors;

[0056] The deduced alarm optimization data is pre-adjusted by compensating the characteristic vector to obtain pre-adjusted alarm optimization data.

[0057] The present invention performs adaptive threshold segmentation on the predicted alarm flooding situation to obtain flooding level data, thereby realizing quantitative evaluation of the alarm flooding degree; performs time segmentation decomposition on the response evaluation data to obtain time response characteristics, thereby capturing the temporal characteristics of the alarm response; performs cross-validation on the flooding level data and the time response characteristics to obtain a response weight matrix, thereby establishing a quantitative relationship between the flooding situation and the response time; performs multi-dimensional decomposition on the deduced alarm optimization data to obtain an optimized feature vector, thereby mining the potential characteristics of the alarm optimization; performs self-organizing mapping on the optimized feature vector according to the response weight matrix to obtain mapping relationship data, thereby optimizing the adjustment strategy of the alarm optimization data; performs compensation condition analysis according to the flooding level data and the time response characteristics to generate compensation condition parameters, thereby accurately analyzing the compensation requirements of the alarm optimization; performs compensation feature conversion on the compensation condition parameters based on the mapping relationship data to generate compensation feature vectors, thereby adjusting the alarm optimization scheme to adapt to different alarm response conditions; performs pre-adjustment on the deduced alarm optimization data through the compensation feature vectors to obtain pre-adjusted alarm optimization data, thereby realizing accurate adjustment of the alarm optimization scheme.

[0058] Preferably, step S6 comprises the following steps:

[0059] Step S61: Perform multi-dimensional statistics on the classified alarm data to obtain statistical dimension data; perform index calculation on the statistical dimension data to obtain classified statistical data;

[0060] Step S62: convert the classified statistical data into a report form to generate an alarm classification statistical report;

[0061] Step S63: Calculate the performance index of the response evaluation data to obtain performance index data; perform benchmark analysis on the performance index data to generate performance benchmark data;

[0062] Step S64: convert the performance benchmarking data into a report form to generate a response performance report;

[0063] Step S65: Based on a preset report template, the alarm classification statistics report, the response performance report and the pre-adjusted alarm optimization data are integrated to generate an alarm analysis report.

[0064] The present invention obtains statistical dimension data by performing multi-dimensional statistics on classified alarm data, accurately quantifies various features of the alarm data, obtains classified statistical data by performing indicator calculation on the statistical dimension data, and provides detailed data for subsequent alarm classification and performance analysis, generates an alarm classification statistical report by performing report conversion on the classified statistical data, clearly displays detailed statistical results of alarm classification, and facilitates subsequent analysis and decision support, obtains performance indicator data by performing performance indicator calculation on response evaluation data, and provides a quantitative basis for evaluating the performance of alarm response, generates performance benchmarking data by performing benchmarking analysis on the performance indicator data, and ensures that the effect of the alarm response is consistent with industry standards or expected goals, generates a response performance report by performing report conversion on the performance benchmarking data, clearly summarizes the performance analysis results, and provides an optimization reference, and performs template integration on the alarm classification statistical report, the response performance report, and the pre-adjusted alarm optimization data based on a preset report template to generate an alarm analysis report, and provides a comprehensive and structured alarm analysis report, which helps to provide a scientific basis for equipment management and optimization.

[0065] The present invention also provides an alarm data analysis system for a power plant, which is used to execute the alarm data analysis method for a power plant as described above. The alarm data analysis system for a power plant comprises:

[0066] The data visualization module is used to collect the original alarm signal; perform frequency calculation on the original alarm signal to obtain the frequency characteristics of the alarm signal; perform time series characteristic analysis on the frequency characteristics of the alarm signal to generate a time series alarm characteristic matrix and perform visualization processing;

[0067] The repeated alarm analysis module is used to locate the repeated alarm points in the time series alarm feature matrix to obtain the repeated alarm points; cluster the alarm types of the repeated alarm points based on the preset evaluation time threshold to generate classified alarm data;

[0068] The alarm correlation analysis module is used to perform inter-class correlation analysis on repeated alarm points based on classified alarm data to obtain alarm point correlation data; perform alarm optimization deduction based on the alarm point correlation data to generate deduced alarm optimization data;

[0069] The alarm flood prediction module is used to obtain alarm history records; perform alarm pattern recognition on the time series alarm feature matrix based on the alarm history records to obtain an alarm pattern sequence; perform flood situation prediction on the alarm pattern sequence to generate a predicted alarm flood situation;

[0070] The alarm response analysis module is used to evaluate the alarm response of the classified alarm data based on the alarm history records and generate response evaluation data; pre-adjust the deduced alarm optimization data based on the predicted alarm flooding situation and response evaluation data to obtain pre-adjusted alarm optimization data;

[0071] The report summary module is used to summarize the classified alarm data, response evaluation data and pre-adjustment alarm optimization data to generate an alarm analysis report.

[0072] The present invention realizes frequency feature extraction and time series feature analysis of the original alarm signal through a data visualization module, generates an alarm feature matrix and presents it visually, and enhances the structuring and intuitiveness of the alarm data. Through a repeated alarm analysis module, the repeated alarm points are located and clustered to generate classified alarm data, providing a data basis for alarm classification and evaluation. Through an alarm correlation analysis module, classified alarm data are used for correlation analysis to generate alarm point correlation data and deduced alarm optimization data, thereby improving the rationality of alarm configuration. Through an alarm flood prediction module, alarm patterns are identified in combination with historical alarm records, and alarm pattern sequences and flood situation prediction results are generated, thereby improving the ability to predict potential alarm problems. Through an alarm response analysis module, alarm response indicators are evaluated and the deduced optimization data are adjusted in combination with flood situation prediction, thereby enhancing the dynamic adaptability of the alarm optimization scheme. Through a report summary module, classified alarm data, response evaluation data and optimization data are integrated to generate an alarm analysis report, thereby providing a comprehensive decision-making basis for power plant alarm management. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] Figure 1 A schematic diagram of the steps of a method for analyzing alarm data of a power plant;

[0074] Figure 2 Detailed implementation flow chart of step S2;

[0075] Figure 3 Detailed implementation flow chart of step S3. DETAILED DESCRIPTION

[0076] The technical method of the present invention is described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by technicians in this field without creative work are within the scope of protection of the present invention.

[0077] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.

[0078] It should be understood that, although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0079] To achieve this, please refer to Figures 1 to 3 , a method for analyzing alarm data of a power plant, comprising the following steps:

[0080] Step S1: collecting the original alarm signal; performing frequency calculation on the original alarm signal to obtain the frequency characteristics of the alarm signal; performing time series characteristic analysis on the frequency characteristics of the alarm signal to generate a time series alarm characteristic matrix, and performing visualization processing;

[0081] Step S2: locating repeated alarm points on the time series alarm feature matrix to obtain repeated alarm points; clustering the alarm types of the repeated alarm points based on a preset evaluation time threshold to generate classified alarm data;

[0082] Step S3: Based on the classified alarm data, inter-class correlation analysis is performed on the repeated alarm points to obtain alarm point correlation data; based on the alarm point correlation data, alarm optimization deduction is performed to generate deduced alarm optimization data;

[0083] Step S4: Acquire alarm history records; perform alarm pattern recognition on the time series alarm feature matrix according to the alarm history records to obtain an alarm pattern sequence; perform flooding situation prediction on the alarm pattern sequence to generate a predicted alarm flooding situation;

[0084] Step S5: performing alarm response evaluation on the classified alarm data based on the alarm history records to generate response evaluation data; pre-adjusting the deduced alarm optimization data based on the predicted alarm flooding situation and the response evaluation data to obtain pre-adjusted alarm optimization data;

[0085] Step S6: Summarize the classified alarm data, response evaluation data and pre-adjustment alarm optimization data to generate an alarm analysis report.

[0086] The present invention can extract the key features of the alarm signal by frequency calculation and time series feature analysis of the original alarm signal. The generated time series alarm feature matrix provides an intuitive data basis for subsequent analysis. The positioning of repeated alarm points and clustering of alarm types realize the effective classification of alarm events. The generated classified alarm data provides support for subsequent correlation analysis. The implementation of inter-class correlation analysis can identify the relationship between alarm points. The obtained alarm point correlation data provides a basis for optimization deduction. The implementation of alarm optimization deduction can generate more accurate optimization strategies. The combination of obtaining alarm history records and alarm pattern recognition can reveal long-term trends. The generated alarm pattern sequence provides a data basis for flooding situation prediction. The generation of response evaluation data can evaluate the effect of alarm response. Alarm response evaluation is performed for optimizing alarm data. Response evaluation data is generated, and the deduction data is adjusted in combination with the prediction results to generate pre-adjustment optimization data. The alarm optimization decision is dynamically adjusted. The classified alarm data, response evaluation data and pre-adjustment optimization data are summarized and analyzed to generate an alarm analysis report, which provides systematic comprehensive evaluation data and guidance basis for equipment alarm management and optimization.

[0087] In the embodiment of the present invention, refer to Figure 1 , is a schematic flow chart of the steps of a method for analyzing alarm data of a power plant according to the present invention. In this example, the method for analyzing alarm data of a power plant comprises the following steps:

[0088] Step S1: collecting the original alarm signal; performing frequency calculation on the original alarm signal to obtain the frequency characteristics of the alarm signal; performing time series characteristic analysis on the frequency characteristics of the alarm signal to generate a time series alarm characteristic matrix, and performing visualization processing;

[0089] In this embodiment, alarm signals from various devices in the power plant are obtained through the real-time monitoring system of the equipment. These signals are usually triggered by sensors or control systems in real time, reflecting the abnormal or warning status of the equipment. The alarm signals include but are not limited to over-limit information of data such as temperature, pressure, flow, vibration, etc. The signal data will be stored on the data server for subsequent processing. Then, the collected alarm signals are frequency calculated using Fourier Transform to obtain the frequency characteristics of the alarm signals. These frequency characteristics are used to analyze the periodicity and frequency of alarm occurrence, and then filter out alarm points with higher or lower frequencies. Next, the frequency characteristics of the alarm signals are analyzed using a time series analysis method to construct a time series alarm feature matrix. This matrix combines time series data with the change characteristics of the alarm signals, and can reveal the time series change rules of the alarm signals. The time series analysis methods used in the analysis process include but are not limited to autoregressive models (AR) and moving average models (MA). The matrix obtained by analysis will be visualized, and the change trend of the alarm signals is displayed through charts to help users intuitively understand the time series characteristics of the alarm signals.

[0090] Step S2: locating repeated alarm points on the time series alarm feature matrix to obtain repeated alarm points; clustering the alarm types of the repeated alarm points based on a preset evaluation time threshold to generate classified alarm data;

[0091] In this embodiment, by calculating the frequency characteristics and time series characteristics of the alarm signal in the matrix, the alarm points that appear repeatedly are identified. The repeated alarm points refer to the points that trigger the alarm multiple times in the same time period. In the identification process, a threshold-based trigger algorithm is used to set a fixed alarm frequency threshold (for example, an alarm point with more than 5 alarm times within an hour is a repeated alarm point). Then, through further analysis of these repeated alarm points, a clustering analysis algorithm is used to classify them. The alarm points are mainly clustered based on the K-means clustering algorithm (K-means clustering) to classify similar alarm points. In the clustering process, the characteristics of each alarm point are set as the input of the clustering to generate classified alarm data. These classified data reflect the occurrence rules and states of different types of alarm points, providing basic data for subsequent alarm analysis.

[0092] Step S3: Based on the classified alarm data, inter-class correlation analysis is performed on the repeated alarm points to obtain alarm point correlation data; based on the alarm point correlation data, alarm optimization deduction is performed to generate deduced alarm optimization data;

[0093] In this embodiment, the correlation coefficient between each type of alarm point is calculated. The commonly used correlation analysis method is the Pearson correlation coefficient. The correlation data between the alarm points are calculated. These correlation data are used to reveal the relationship between different types of alarm points, especially when certain alarm points occur at the same time, whether they have a strong dependence on other alarm points. Next, based on the alarm point correlation data, alarm optimization deduction is performed. The specific operation is to use a model-based deduction algorithm, combine historical alarm data and existing alarm point correlations, deduce optimized alarm configuration parameters, and generate deduced alarm optimization data. This data will provide a reference for subsequent alarm optimization strategies. Model selection in the deduction process includes but is not limited to deduction models based on regression analysis and machine learning algorithms.

[0094] Step S4: Acquire alarm history records; perform alarm pattern recognition on the time series alarm feature matrix according to the alarm history records to obtain an alarm pattern sequence; perform flooding situation prediction on the alarm pattern sequence to generate a predicted alarm flooding situation;

[0095] In this embodiment, all alarm data in the past period of time are extracted from the database, including information such as the alarm type, time, equipment, and response time. The data is organized into a standardized format for analysis. Next, based on these historical records, the time series alarm feature matrix is ​​subjected to alarm pattern recognition. By matching the alarm patterns in the historical records, a pattern recognition method based on neural networks is used to classify and identify different types of alarm patterns to generate an alarm pattern sequence. During the alarm pattern recognition process, a long short-term memory network (LSTM) is used for deep learning, which can effectively capture the long-term dependencies in the time series data. Next, the alarm pattern sequence is subjected to flooding trend prediction. By statistically analyzing the changes in the alarm patterns in the past and combining the alarm frequency and time in the historical data, a predicted alarm flooding trend is generated. In the prediction process, a sliding window method is used to set the alarm frequency within a certain time period as the prediction standard to evaluate whether the alarm has entered a flooding state.

[0096] Step S5: performing alarm response evaluation on the classified alarm data based on the alarm history records to generate response evaluation data; pre-adjusting the deduced alarm optimization data based on the predicted alarm flooding situation and the response evaluation data to obtain pre-adjusted alarm optimization data;

[0097] In this embodiment, alarm response data in historical records are extracted, including alarm confirmation time, recovery time, etc., and the response time is calculated based on the set indicator formula to generate response evaluation data. Next, combined with the predicted alarm flooding situation, the response evaluation data is used to pre-adjust the deduced alarm optimization data. The goal of the adjustment is to make the alarm optimization plan more in line with the actual alarm response situation. An adjustment method based on a genetic algorithm is used in the optimization process. The optimization parameters are adjusted through continuous iteration until the optimal alarm optimization data is obtained. Finally, the pre-adjusted alarm optimization data is output to provide a basis for the system's alarm strategy adjustment.

[0098] Step S6: Summarize the classified alarm data, response evaluation data and pre-adjustment alarm optimization data to generate an alarm analysis report.

[0099] In this embodiment, when the classified alarm data, response evaluation data and pre-adjustment alarm optimization data are aggregated, each type of data is first integrated according to a specific format, the classified alarm data and the response evaluation data are associated through database query, and the pre-adjustment alarm optimization data is generated through the processing results of the aforementioned steps. Next, a report generation tool is used to format these aggregated data and convert them into a standard alarm analysis report. The generated report includes key indicators such as alarm frequency, type distribution, and response time analysis. Finally, all contents are summarized in an alarm analysis report to facilitate operators and management to make further decisions. During the report generation process, a template-based automated report generation system is used to automatically fill in data and generate a final report by defining a standard report template.

[0100] Preferably, step S1 comprises the following steps:

[0101] Step S11: collecting the original alarm signal; performing digital filtering on the original alarm signal to obtain a filtered alarm signal;

[0102] Step S12: segmenting the filtered alarm signal into time windows to obtain time period alarm data; performing overlapping window processing on the time period alarm data to obtain overlapping time period alarm data;

[0103] Step S13: Perform frequency statistics processing on the alarm data of the overlapping period to generate the frequency characteristics of the alarm signal; perform time series correlation calculation on the frequency characteristics of the alarm signal to obtain time series characteristic data;

[0104] Step S14: performing matrix conversion processing on the time series feature data to obtain a time series alarm feature matrix; and visually displaying the time series alarm feature matrix.

[0105] In this embodiment, the alarm signal of the equipment is obtained from the automatic monitoring system (SCADA) of the power plant. The original alarm signal data includes the status information of the equipment, the alarm value of the sensor, etc. These signals record the working status of the equipment and are triggered when the equipment is abnormal. The alarm signal includes the temperature, pressure, current and other excessive data of the equipment. After the data of the alarm signal is transmitted to the data processing server, the original alarm signal is digitally filtered using a low-pass filter, and a cutoff frequency (for example, 10Hz) is set to remove high-frequency noise. This filter removes unnecessary frequency components in the original signal and retains useful low-frequency components. The signal after filtering is the filtered alarm signal. The filtering operation adopts the finite impulse response (FIR, Finite Impulse Response) filter, the filtered signal removes unnecessary fluctuations caused by electrical noise, signal interference and other factors, and performs time window segmentation processing on it, dividing the continuous alarm signal data into several fixed time periods (for example, every 10 minutes is a period), and the data of each period contains all the alarm signals in the time period. The division of this time period is based on the alarm recording time of the equipment, and the length of the time period can be adjusted according to actual needs. Next, the time period alarm data is overlapped The window processing is set with an overlap coefficient (for example, 50%), and on the basis of each time period, the sliding window method is used to move the window forward by half a time period (for example, every 5 minutes) to generate new overlapping time period alarm data. The purpose of this processing method is to increase the coverage of alarm data and improve the timeliness and accuracy of data. The generation of overlapping windows is realized by programming, and the data structure of the sliding window adopts a ring buffer (Ring Buffer) to ensure the continuity and non-loss of each data. The overlapping period alarm data finally obtained is a complete data set extracted from multiple time periods. The frequency of occurrence of various alarm signals in each time period is counted. The time period is set to 10 minutes. The alarm type (such as high temperature, low pressure, etc.) and its occurrence frequency are counted every 10 minutes. The frequency statistics are completed by using a counter. During the statistical process, the alarm type and the number of occurrences are associated to obtain the frequency of occurrence of each alarm type. Subsequently, the time series correlation calculation is performed on the frequency characteristics of the alarm signals obtained by statistics. The Pearson correlation coefficient is used to perform correlation analysis on the frequency of alarm signals in each time period to obtain the correlation data of different alarm types in different time periods. The purpose of this step is to identify the relationship between alarm types, such as whether certain alarm types occur simultaneously in the same time period regularly. In the process of calculating the time series correlation, the frequency characteristics of each alarm signal are taken as input, and the correlation coefficient matrix is ​​calculated in combination with data analysis tools (such as the NumPy library in Python).After obtaining the time series feature data, perform matrix conversion on it. The original time series feature data is constructed into a matrix according to the time sequence and alarm signal type. The rows of the matrix represent time periods, the columns represent different types of alarm signals, and the value of each cell is the frequency value of the corresponding alarm signal in the time period. The matrix conversion uses data processing tools (such as the DataFrame structure in the Pandas library). The converted matrix contains the frequency feature information of all alarm signals in the entire analysis period. Next, the time series alarm feature matrix is ​​visualized by using data visualization tools such as Matplotlib or Plotly to convert the matrix into a chart, such as a heat map (Heatmap) or a line chart (Line Chart). The chart can intuitively display the changes in the alarm signal in different time periods. The color depth in the heat map represents the intensity of the alarm signal. The changes on the time axis reflect the temporal change law of the alarm signal. Finally, the visualization result is output as a chart file. ,

[0106] Preferably, step S2 comprises the following steps:

[0107] Step S21: Perform frequency statistical analysis on the time series alarm feature matrix to obtain alarm frequency data;

[0108] Step S22: performing threshold comparison and screening on the alarm frequency data according to a preset alarm frequency threshold to obtain repeated alarm data; locating repeated alarm points according to the repeated alarm data to obtain repeated alarm points;

[0109] Step S23: performing time dimension aggregation on repeated alarm points to generate time clustering data; performing feature focusing on the time clustering data to obtain feature clustering data;

[0110] Step S24: classify the feature clustering data by type labeling to generate classified alarm data.

[0111] In this embodiment, the frequency of occurrence of each alarm signal in the time series alarm feature matrix in each time period is counted. The statistical process uses a frequency analysis algorithm. For each column (representing a certain type of alarm signal), all rows (representing different time periods) are traversed to calculate the frequency of occurrence of each alarm signal in the time period. During the calculation process, the number of all alarm signals that appear is accumulated to finally obtain the frequency data of each alarm signal. These frequency data reflect the occurrence of alarm signals in different time periods. After the statistics are completed, the frequency data is stored through a data frame (DataFrame). The frequency information of each alarm signal corresponds to its corresponding alarm type one by one. An alarm frequency threshold is defined (for example, it is set that the alarm signal is a repeated alarm signal if it appears more than 3 times within a time period of 60 seconds). The alarm frequency data is compared to filter out data points with a frequency greater than the threshold. Condition-based filtering is used during the screening. The method of filtering is used to traverse the alarm frequency data, compare whether the frequency value of each signal is greater than the preset threshold, and filter out the alarm signals that meet the conditions. After the screening is completed, a group of repeated alarm data is obtained. Next, the repeated alarm points are located according to the repeated alarm data. When locating, the time period of repeated alarm signals is found to mark the time points when alarm signals frequently appear. The location of repeated alarm points uses timestamp data to match the frequency data of alarm signals, mark the time points when multiple alarm signals appear, and obtain repeated alarm points. The repeated alarm points are aggregated according to the time dimension. The aggregated time unit can be hours, days or weeks, which is adjusted according to the time interval of the alarm signal. During the aggregation process, the repeated alarm points are grouped by timestamps, and all repeated alarm points in the same time period are merged into one group. The clustering algorithm is used. Algorithm) analyzes the repeated alarm points within the time period to generate time clustering data. The time clustering data includes all the repeated alarm points that occur within the same time period. The clustering operation is completed by calculating the time difference between the repeated alarm points, and the similar alarm points are classified into one category. The threshold of the time difference can be set to a certain time range (such as alarm signals within 5 minutes are regarded as the same category). The time clustering data is stored in a data structure (such as an array or a list), and the storage format is a time period and its corresponding set of repeated alarm points. Next, the time clustering data is feature focused. Feature focusing extracts the type, frequency and other information of the alarm signal in each time period, and weights and merges these features to obtain feature clustering data. During the feature focusing operation, the frequency distribution, alarm type and other features of the clustered repeated alarm point data are calculated, and several alarm signal types (such as high temperature, low voltage, current overload, etc.) are defined. Then, the alarm signal features in the feature clustering data are marked as types.The labeling is realized through the classification algorithm. The alarm signals in each cluster are matched according to their characteristics (such as frequency, alarm value, etc.), the most appropriate alarm type is selected and labeled. The classification process uses a machine learning classification model (such as a decision tree classifier) ​​to train according to the feature data to generate a trained classification model. The model is used to classify the feature cluster data, and the data of each cluster is classified into the corresponding alarm type to obtain the classified alarm data. The classified alarm data contains each alarm type and its corresponding alarm signal information. Finally, the classification results are displayed through visualization tools (such as Python's Matplotlib). ,

[0112] Preferably, step S3 comprises the following steps:

[0113] Step S31: constructing a feature vector for the classified alarm data to obtain feature description data; performing similarity calculation on the feature description data to generate a similarity matrix;

[0114] Step S32: performing inter-class correlation analysis on repeated alarm points based on the similarity matrix to obtain alarm point correlation data;

[0115] Step S33: performing optimization rule matching on the alarm point correlation data to obtain optimization strategy data; performing numerical simulation test based on the optimization strategy data to generate numerical simulation results;

[0116] Step S34: Perform alarm optimization deduction on the optimization strategy data based on the numerical simulation results to generate deduced alarm optimization data.

[0117] In this embodiment, key features are extracted from each classified alarm data. These features include the time, frequency, alarm level, fluctuation amplitude of the alarm signal, etc. The feature vector of each alarm data is constructed through these features. The feature vector is constructed by using a standardization method (Standardization). Each feature is normalized to ensure that the dimensions of different features are consistent. The obtained feature vector is a numerical data set. Each alarm data corresponds to a feature vector. The feature vector contains various aspects of the alarm signal. The feature description data is represented by a constructed vector matrix (Feature Matrix). The feature description data is a multi-dimensional data representation of the alarm signal. Then, the feature description data is similarity calculated. The cosine similarity algorithm (Cosine Similarity) is used to calculate the similarity between the feature vectors. Specifically, the similarity value is calculated by the dot product and the vector modulus to generate a similarity matrix (Similarity Matrix), which records the similarity between each pair of alarm signals. Each element of the matrix represents the similarity between two alarm signals in the feature vector space, and finally obtains the similarity matrix. The similarity matrix is ​​used to calculate the correlation between each type of alarm signals. The inter-class correlation analysis calculates the similarity of each pair of alarm signals and determines which alarm signals belong to the same category by combining the preset correlation threshold. The specific operation is to perform cluster analysis on the similarity values ​​in the matrix and use the K-means clustering algorithm to divide the alarm signals into different clusters according to their similarities. Each cluster contains similar alarm signals. In the clustering process, the similarity between signals is quantified by setting a distance metric (such as Euclidean distance). The inter-class correlation data is recorded through the center point of the clustering result and the similarity value of each alarm signal in the cluster. The recorded data is recorded in the data table (Data Table). Table), each row represents a cluster, including similarity information of alarm points in the cluster and central features of the cluster. According to the inter-class correlation data obtained in step S32, optimization rules are set. These rules are based on the correlation between alarm points. For example, when two alarm signals exceed a certain threshold in similarity, they can be identified as redundant alarm signals, and the alarm levels can be merged or adjusted. The matching of optimization rules is performed through logical operations. The threshold settings involved in the rules include similarity value thresholds, time difference thresholds, etc. The alarm signals that need to be optimized are screened out through these rules to obtain optimization strategy data. The optimization strategy data includes the adjustment method of each alarm signal, such as merging, deleting or recalibrating. The optimization strategy data is stored in a data frame and provides a corresponding optimization operation for each alarm signal. When numerical simulation tests are performed based on the optimization strategy data,Using the simulation model, by inputting the optimized alarm signal data, simulating the response behavior of the alarm signal, and generating numerical simulation results, the numerical method (such as the finite difference method) is used to model the timing characteristics of the alarm signal during the simulation process, and the numerical simulation result data is loaded. These data include the behavior pattern of the optimized alarm signal in the simulation environment. Then, according to the simulation results, the optimization strategy is further deduced through the preset deduction algorithm. During the deduction process, the performance of the optimized alarm signal in the actual system is analyzed based on the simulation data to determine whether it can effectively reduce redundant alarms or improve the alarm response speed. During the deduction, the deduction model is used to infer the optimization strategy data. According to the feedback data in the deduction process, the details of the optimization strategy are adjusted until the final alarm optimization strategy is obtained. The deduction alarm optimization data records the performance of the optimization strategy under different test conditions. Finally, the deduction alarm optimization data is displayed through visualization tools (such as Python's Matplotlib or Seaborn). ,

[0118] Preferably, step S32 includes the following steps:

[0119] Based on the similarity matrix, the repeated alarm points are associated and mapped to obtain the alarm point dependency matrix; based on the alarm point dependency matrix, the repeated alarm points are associated and analyzed to generate the alarm point associated trajectory;

[0120] Perform aggregation analysis based on alarm point correlation trajectories to obtain alarm point behavior patterns; extract inter-point behavior rules from alarm point behavior patterns to obtain inter-point behavior rules;

[0121] The separation degree of repeated alarm points is calculated based on the behavior rules between points to generate the inter-class sparsity;

[0122] A multi-layer correlation network is constructed according to the inter-class sparsity and alarm point dependency matrix to obtain the alarm point association network; the alarm point association network is analyzed layer by layer to generate alarm point correlation data.

[0123] In this embodiment, similarity data between alarm points are extracted from the similarity matrix, and the dependency relationship between each pair of alarm points is calculated by using a matrix operation method. A dependency threshold is set. If the similarity between two alarm points exceeds the threshold, it is considered that they have a dependency relationship, and an alarm point dependency matrix is ​​generated. The matrix is ​​a two-dimensional array. The elements in the matrix represent the dependency strength between the alarm points, and the dependency strength is represented by the similarity value. Then, the association trajectory of the repeated alarm points is analyzed based on the alarm point dependency matrix. By traversing each pair of alarm points in the matrix, the temporal changes of the alarm points are tracked, and the association trajectory of the alarm points is constructed. The association trajectory records the dependency relationship and the change law between the alarm points in the time dimension, and finally the alarm point association trajectory is generated. According to the time information in the alarm point association trajectory, the alarm point association trajectory is aggregated. The aggregation analysis identifies the main behavior pattern by combining similar trajectories, and the similar trajectories are classified using a clustering algorithm (such as the DBSCAN clustering algorithm) to generate the alarm point behavior pattern. The alarm point behavior pattern represents the behavior characteristics between different alarm points in the same time period. The behavior pattern includes the frequency, duration, and strength of the alarm signal. By aggregating the associated trajectories of multiple alarm points, multiple behavior patterns are obtained. Each behavior pattern corresponds to a feature set of an alarm point. Next, the behavior patterns between alarm points are extracted based on the behavior patterns. The behavior patterns between alarm points are extracted by calculating the similarity between the behavior patterns to extract the behavior patterns between points. The specific operation is to determine the behavior patterns between alarm points by calculating the intra-class similarity in the trajectory clustering. According to the behavior patterns between points, the behavior differences between alarm points are calculated. By calculating the difference in the behavior pattern of each alarm point, the separation degree between alarm points is obtained. The separation degree calculation is measured by measuring the similarity difference of the behavior patterns, and using statistical methods (such as standard deviation Standard Deviation) calculates the degree of separation between the behavior patterns of each pair of alarm points to obtain the inter-class sparsity. The inter-class sparsity value reflects the independence between the alarm points. A higher inter-class sparsity indicates that the behaviors of the alarm points are very different, and a lower inter-class sparsity indicates that the behaviors of the alarm points are similar. The sparsity data is stored as a matrix, in which each element represents the inter-class sparsity value between the alarm points. Then, a multi-layer correlation network is built based on the inter-class sparsity and the alarm point dependency matrix. First, by weighting the inter-class sparsity data and setting the weight coefficient, a multi-level alarm point association network is constructed in combination with the alarm point dependency matrix. In this network, each alarm point is used as a node, and the edges between the nodes represent the dependency relationship and behavior rules between the alarm points. Through the hierarchical network structure,A multi-level alarm point dependency network is formed. Then, when the alarm point association network is parsed layer by layer, each layer of the network is traversed to extract the dependency relationship between the nodes in each layer, and the node relationship of each layer is gradually parsed to finally obtain the correlation data of the alarm point.

[0124] Preferably, step S4 comprises the following steps:

[0125] Step S41: Acquire alarm history records; perform time completion processing on the alarm history records to obtain completed history records;

[0126] Step S42: performing periodic analysis on the time series alarm feature matrix based on the completed historical records to generate periodic feature data; performing alarm trend recognition on the periodic feature data to obtain trend feature data;

[0127] Step S43: performing pattern decomposition on the trend feature data to obtain pattern component data; performing pattern matching on the pattern component data and the time series alarm feature matrix to generate an alarm pattern sequence;

[0128] Step S44: Decompose the alarm pattern sequence to obtain pattern sequence features; perform time series change trend analysis on the pattern sequence features to generate alarm pattern change trends;

[0129] Step S45: Perform situation assessment based on the alarm mode change trend to generate a predicted alarm flooding situation.

[0130] In this embodiment, alarm data within a certain time range in the past are extracted from the monitoring system of the power plant. These alarm data include information such as the time when the alarm occurred, the alarm type, the alarm point number, the alarm level, and the duration of the alarm. A complete historical data set is extracted from the alarm record table using a database query tool (such as SQL query) to obtain an alarm history record. Then, the alarm history record is time-completed. First, it is detected whether there are time gaps or abnormal values ​​in the history record. An interpolation method (such as linear interpolation) is used to complete the time gap. For the missing alarm time, a reasonable alarm timestamp is calculated by interpolation and inserted into the original data to ensure the integrity of the data in the time dimension. A completed history record is obtained. The data includes continuous alarm records with uniform time intervals. The time series of the data is repaired and supplemented. A time series alarm feature matrix is ​​established. The rows of the matrix represent alarm points, the columns represent time, and each element in the matrix represents the alarm level or alarm state at a specific time point. The Fast Fourier Transform (FFT) is used to calculate the time series of the alarm records. Transform) is used to periodically analyze the time series alarm feature matrix, analyze the periodic law of the alarm data on the time axis, and convert the time domain data into frequency domain data through Fourier transform, extract the periodic characteristics of the alarm signal, and obtain periodic characteristic data. This data reflects the regular changes of different alarm points in different time periods. Then, the alarm trend is identified for the periodic characteristic data. The time series data trend detection method (such as Holt-Winters smoothing method) is used to analyze the periodic characteristic data to identify the rising, falling or stable trend in the alarm data. Finally, the trend characteristic data is obtained, which records the trend changes of the alarm points in the future time period. The principal component analysis (PCA) is used to reduce the dimension of the trend characteristic data, reduce the high-dimensional alarm trend data to a low dimension, extract the most representative pattern components, and obtain the pattern component data. The pattern component data includes the main change patterns of each alarm point, representing the core trend and behavior of the alarm data. Then, the pattern matching is performed on the pattern component data and the time series alarm feature matrix, and dynamic time warping (DTW) is used to perform pattern matching on the pattern component data and the time series alarm feature matrix. The Alarm Warping algorithm is used to match the pattern component data with the time series alarm feature matrix, calculate the similarity between the alarm points, find the pattern sequence in the historical alarm data that is most similar to the current alarm pattern, and generate an alarm pattern sequence. The sequence includes the pattern prediction of future alarm points, and performs time series analysis on each alarm pattern in the alarm pattern sequence. The pattern sequence is decomposed by applying wavelet transform to obtain pattern features of different frequency bands. The wavelet transform can extract the change features of the alarm pattern at multiple scales from low frequency to high frequency, and obtain the pattern sequence features.The pattern sequence characteristics describe the changing rules of the alarm pattern at various time scales.

[0131] Next, the time series change trend of the pattern sequence characteristics is analyzed, and the time series analysis method (such as autoregressive model AR, AutoRegressive) is used to model the pattern sequence characteristics, analyze the change trend of the pattern sequence in the future time period, generate the alarm pattern change trend, define the situation assessment model, and establish the association model between the alarm pattern change trend and the alarm risk through the analysis of historical data. The risk assessment method (such as Bayesian Network Bayesian Network) is used to evaluate the alarm pattern change trend, generate risk prediction data, and evaluate the alarm frequency and severity of each alarm point in the future period. Then, the model is used to calculate the alarm risk and generate a predicted alarm flood situation. The data of the predicted alarm flood situation is calculated based on the risk prediction data, and the alarm flood area and its alarm intensity are output. Further, based on these data, a preventive maintenance plan for power plant equipment is arranged to reduce potential safety hazards caused by alarm flooding.

[0132] Preferably, step S5 comprises the following steps:

[0133] Step S51: Analyze the response time of the alarm history records to obtain historical response time data;

[0134] Step S52: mapping the classified alarm data to a response time according to the historical response time data to generate a predicted current alarm response time;

[0135] Step S53: Use the preset index and alarm scoring formula to perform alarm response evaluation on the predicted current alarm response time to generate response evaluation data, wherein the alarm scoring formula is:

[0136]

[0137] Among them, S is the final score, V is the achieved value of the device, and T is the index of this value;

[0138] Step S54: self-organizing mapping is performed on the predicted alarm flooding situation and response evaluation data to obtain mapping relationship data; and pre-adjusting the deduced alarm optimization data based on the mapping relationship data to obtain pre-adjusted alarm optimization data.

[0139] In this embodiment, alarm record data within a certain period of time in the past is extracted from the alarm data management system of the power plant, and the data includes the trigger time, response time and response processing time of each alarm record. For each alarm record, the response time, that is, the time interval between alarm triggering and response, is calculated. A data analysis tool (such as the Pandas library in Python) is used to clean and organize all alarm records, and invalid or missing response data are eliminated to obtain a response time data set, in which each record includes the alarm trigger time, response start time, response end time and response time. Statistical indicators of the response time of the data set, such as the mean, standard deviation, etc., are further calculated to finally obtain complete historical response time data, which describes the response efficiency and response time distribution of each alarm point in the past period of time. According to the historical response time data, the response time of the alarm point is classified according to the alarm type, such as equipment failure type, environmental abnormality type, etc., and statistical methods (such as K-means clustering K-means) are used. The alarm types are grouped by clustering to obtain the distribution of alarm response time of different categories. Then, the current alarm data to be processed is compared with the historical classification data to extract the category of the current alarm. The predicted response time of the current alarm is obtained by calculating the average and standard deviation of the response time of the current alarm category and similar categories in the historical data. The response time of the historical data is mapped to the current alarm point using an interpolation method (such as weighted interpolation based on similarity), thereby predicting the response time of the current alarm and generating the predicted current alarm response time data. Evaluation indicators are set, including device type, alarm level, response time, etc. Then, the predicted current alarm response time is evaluated using the alarm scoring formula, which is: Among them, S is the alarm response score, V is the current predicted response time, and T is the preset standard value of the response time of the alarm point. This formula is used to calculate each alarm data. According to the parameters in the formula, the alarm response score of each record is calculated. If V (response time) is greater than T (preset standard value), the score will be lower. If V is close to T, the score will be higher. Finally, a set of response evaluation data is obtained, which records the response score of each alarm point. The self-organizing map network (SOM) is used to analyze the predicted alarm flooding situation data. The SOM algorithm clusters the alarm data through unsupervised learning, classifies similar alarm modes into one category, and obtains mapping relationship data. The mapping relationship data reflects the relationship between the potential trend of alarm flooding and the response time. Then, based on the mapping relationship data, the deduced alarm optimization data is pre-adjusted, and the optimization algorithm (such as particle swarm optimization PSO, Particle Swarm Optimization PSO) is applied. The deduced alarm optimization data includes adjustments to the alarm priority, processing time, and preventive measures of the equipment. By analyzing the alarm flooding situation and response evaluation data, the alarm response strategy in the deduced data is adjusted to obtain the pre-adjusted alarm optimization data. This data reflects the alarm management plan optimized according to the actual alarm data, and provides a decision-making basis for subsequent alarm processing.

[0140] Preferably, step S54 includes the following steps:

[0141] Adaptively segment the predicted alarm flooding situation by thresholds to obtain flooding level data; decompose the response evaluation data into time segments to obtain time response characteristics;

[0142] Cross-validate the flooding level data and time response characteristics to obtain the response weight matrix;

[0143] Perform multi-dimensional decomposition on the deduction alarm optimization data to obtain the optimized feature vector; perform self-organizing mapping on the optimized feature vector according to the response weight matrix to obtain mapping relationship data;

[0144] Compensation condition analysis is performed based on flooding level data and time response characteristics to generate compensation condition parameters; compensation characteristic conversion is performed on the compensation condition parameters based on mapping relationship data to generate compensation characteristic vectors;

[0145] The deduced alarm optimization data is pre-adjusted by compensating the characteristic vector to obtain pre-adjusted alarm optimization data.

[0146] In this embodiment, data on the predicted alarm flooding situation are collected. These data include indicators such as alarm triggering frequency, alarm duration, and response delay. Then, based on these data, an adaptive threshold segmentation algorithm is applied to segment the alarm flooding situation data according to different thresholds. The threshold is dynamically determined by a statistical method (such as a quantile method). According to changes in the alarm frequency and response time, the segmentation point is dynamically adjusted to divide the alarm flooding situation into different flooding levels. The flooding level data will reflect the severity of the alarm mode, such as being divided into three levels: low, medium, and high, and the corresponding flooding level of each alarm is recorded. The flooding level data finally obtained includes different intensities of alarms and potential threat levels. The response evaluation data is divided into multiple time periods (such as every hour, every day, etc.) according to time. Then, according to the response situation in each time period, the characteristic values ​​such as the mean, maximum, and minimum value of the response time in the time period are calculated. Then, these response data are analyzed using a time series analysis method (such as a sliding window method). The first step is fragmented decomposition. Each time segment reflects the response situation within a certain period of time. The obtained time response characteristic data will include the dynamic changes of the response within each time period, such as the peak value, fluctuation amplitude, response stability, etc. of the response time. Finally, the time response characteristic data is obtained. The flooding level data and the time response characteristic data are combined into a composite data set. Then, the composite data set is verified using a cross-validation method (such as K-fold cross-validation). The specific method is to divide the data set into multiple subsets, train the model on each subset and test the generalization ability of the model. Finally, according to the results of cross-validation, the influence weight of each feature (flooding level and time response feature) on the alarm response is calculated, and the response weight matrix is ​​obtained by weighted average method. Each value in the response weight matrix represents the importance of flooding level or time response feature in response evaluation. The deduced alarm data, including alarm type, response time, equipment status and other information, is collected and deduced. Then, principal component analysis (PCA, Principal Component Analysis) method is used to reduce the dimension of the data and extract the main feature dimensions in the data. Each dimension represents different aspects of the alarm data that affect alarm optimization, such as the impact of alarm type, the impact of response time, etc. After decomposition, the optimized feature vector is obtained. Each vector is composed of multiple principal components, and each principal component represents an important type of alarm feature. Finally, the optimized feature vector is generated. When the self-organizing map (SOM) algorithm is used to process the optimized feature vector, first, the optimized feature vector is input into the self-organizing map algorithm, and the network structure and training parameters of the self-organizing map are set. Then, each optimized feature vector is weighted according to the response weight matrix. Feature vectors with higher weights will occupy more representative positions in the self-organizing map process. Then, the algorithm generates a low-dimensional mapping result through repeated training and competitive learning.The mapping relationship data reflects the distribution of the optimized features in different alarm modes. Through self-organizing mapping, the mapping relationship data obtained reveals the potential connection between the optimized features and the alarm mode. Using the known flooding level data and time response feature data, a compensation condition model is established through regression analysis methods (such as linear regression or decision tree regression). The model is used to analyze the relationship between the flooding level and the response time. Then, based on the model analysis results, the compensation conditions of the response characteristics under each flooding level are determined. These compensation conditions include the adjustment value of the alarm response time, the optimized alarm processing sequence, etc. Finally, the compensation condition parameters are obtained and the mapping relationship data are collected. The data reflects the feature distribution of different alarm modes and the corresponding optimization results. Then, the features in the mapping relationship data are used to compensate for the The condition parameters are converted. The conversion method can use interpolation or mapping function to map the compensation condition parameters to the optimization feature space to obtain the compensation feature vector. The compensation feature vector is composed of different compensation conditions, and each condition corresponds to a different alarm response adjustment value. Finally, the compensation feature vector obtained contains all the necessary adjustment information. The compensation conditions in the compensation feature vector are used to adjust the various data in the deduced alarm optimization data. The specific adjustments include corrections to the alarm response time, priority, alarm processing strategy, etc. Then, the alarm optimization model is preprocessed using the adjusted data to ensure that the alarm optimization data can more accurately reflect the actual alarm situation during the implementation process. Finally, the pre-adjusted alarm optimization data obtained contains the alarm optimization plan adjusted based on the compensation conditions.

[0147] Preferably, step S6 comprises the following steps:

[0148] Step S61: Perform multi-dimensional statistics on the classified alarm data to obtain statistical dimension data; perform index calculation on the statistical dimension data to obtain classified statistical data;

[0149] Step S62: convert the classified statistical data into a report form to generate an alarm classification statistical report;

[0150] Step S63: Calculate the performance index of the response evaluation data to obtain performance index data; perform benchmark analysis on the performance index data to generate performance benchmark data;

[0151] Step S64: convert the performance benchmarking data into a report form to generate a response performance report;

[0152] Step S65: Based on a preset report template, the alarm classification statistics report, the response performance report and the pre-adjusted alarm optimization data are integrated to generate an alarm analysis report.

[0153] In this embodiment, classified alarm data are collected, which include information such as alarm type, alarm frequency, device status, response time, etc. Then, multi-dimensional statistics are performed on these data, and multiple dimensions (such as alarm frequency, device failure type, alarm duration, alarm processing time, etc.) are selected for analysis. table) and other tools to summarize and process the data, calculate the alarm frequency, mean, standard deviation and other statistics on each dimension, and obtain statistical dimension data. These data show the distribution of different alarm types in each dimension, such as the alarm frequency of equipment failure type, the average response time, etc., to obtain statistical dimension data. Based on the statistical dimension data, select key indicators such as alarm frequency and response time for calculation, and use weighted average method or weighted standard deviation method to calculate the key statistical indicators of each dimension. Through these calculations, obtain classified statistical data, which shows the performance of each alarm type in each dimension. For example, the average response time of a certain type of alarm is 10 minutes, the alarm frequency of equipment failure type is 60 times, etc., to obtain classified statistical data. Based on the classified statistical data, select the data fields to be displayed, such as alarm type, frequency, response time, processing time, etc. Then, use Excel, Pandas library (data analysis library in Python) or special report generation tools to format these data, and display each alarm category and its Statistical indicators are displayed in the report, and the header and table style are set to facilitate reading and understanding. Then, the calculated classification statistical data are filled into the report template, and the data is sorted and summarized as needed to generate an alarm classification statistical report. The report can include important data such as the frequency and processing time of different alarm types. The alarm classification statistical report is obtained, and response evaluation data is collected, including alarm response time, response processing speed, equipment status change and other data. Then, performance indicators are selected, such as average response time, maximum response time, minimum response time, response stability, etc., and these indicators are calculated using standard statistical methods (such as mean, standard deviation, extreme value analysis, etc.) to obtain performance indicator data. These indicator data can effectively reflect the performance of the response system when processing alarms. For example, the average response time is calculated to be 3 minutes, and the response fluctuation range is 1 minute. The final performance indicator data includes all key performance indicators. Known performance standards or historical performance data in the industry or power plant are collected. Then, benchmarking analysis is used. Analysis) compares the current performance indicator data with these standards to analyze whether the response performance of the current system meets the expected standards or the industry average. Performance benchmarking data is obtained through comparative analysis. For example, the average response time is compared with the industry standard. It is found that the response time of the current system is 1 minute higher than the industry standard. The performance benchmarking data is extracted from the benchmarking analysis results. These data include comparison results with industry standards, performance gaps, and optimization suggestions.Next, use report generation tools (such as Excel, Word, etc.) to organize these data into reports, set the report title, table structure, data fields, etc., fill the performance gap, response time comparison value and other indicators into the report, then adjust the visual layout of the table as needed to make it more intuitive and easy to read, and finally generate a response performance report, integrate the generated alarm classification statistics report and response performance report according to the preset template format, the template format includes the structure, title, chapter content, etc. of the report, then fill in the key data, charts, and analysis results in each report according to the report template requirements to ensure that the data display and content description conform to the standard format, then integrate the pre-adjustment alarm optimization data, and display the data as an appendix or part of the report at the end of the report. The report template requires that all data be presented in a clear format, such as sorting by priority, grouping by category, etc. Finally, generate an alarm analysis report, which includes alarm classification statistics, response performance benchmarking data, and pre-adjustment alarm optimization data. The report format is consistent and the structure is clear, which is convenient for managers to quickly understand the alarm data analysis results. The report is finally output as a PDF or Word document for easy circulation and archiving. ,

[0154] The present invention also provides an alarm data analysis system for a power plant, which is used to execute the alarm data analysis method for a power plant as described above. The alarm data analysis system for a power plant comprises:

[0155] The data visualization module is used to collect the original alarm signal; perform frequency calculation on the original alarm signal to obtain the frequency characteristics of the alarm signal; perform time series characteristic analysis on the frequency characteristics of the alarm signal to generate a time series alarm characteristic matrix and perform visualization processing;

[0156] The repeated alarm analysis module is used to locate the repeated alarm points in the time series alarm feature matrix to obtain the repeated alarm points; cluster the alarm types of the repeated alarm points based on the preset evaluation time threshold to generate classified alarm data;

[0157] The alarm correlation analysis module is used to perform inter-class correlation analysis on repeated alarm points based on classified alarm data to obtain alarm point correlation data; perform alarm optimization deduction based on the alarm point correlation data to generate deduced alarm optimization data;

[0158] The alarm flood prediction module is used to obtain alarm history records; perform alarm pattern recognition on the time series alarm feature matrix based on the alarm history records to obtain an alarm pattern sequence; perform flood situation prediction on the alarm pattern sequence to generate a predicted alarm flood situation;

[0159] The alarm response analysis module is used to evaluate the alarm response of the classified alarm data based on the alarm history records and generate response evaluation data; pre-adjust the deduced alarm optimization data based on the predicted alarm flooding situation and response evaluation data to obtain pre-adjusted alarm optimization data;

[0160] The report summary module is used to summarize the classified alarm data, response evaluation data and pre-adjustment alarm optimization data to generate an alarm analysis report.

[0161] The present invention realizes frequency feature extraction and time series feature analysis of the original alarm signal through a data visualization module, generates an alarm feature matrix and presents it visually, and enhances the structuring and intuitiveness of the alarm data. Through a repeated alarm analysis module, the repeated alarm points are located and clustered to generate classified alarm data, providing a data basis for alarm classification and evaluation. Through an alarm correlation analysis module, classified alarm data are used for correlation analysis to generate alarm point correlation data and deduced alarm optimization data, thereby improving the rationality of alarm configuration. Through an alarm flood prediction module, alarm patterns are identified in combination with historical alarm records, and alarm pattern sequences and flood situation prediction results are generated, thereby improving the ability to predict potential alarm problems. Through an alarm response analysis module, alarm response indicators are evaluated and the deduced optimization data are adjusted in combination with flood situation prediction, thereby enhancing the dynamic adaptability of the alarm optimization scheme. Through a report summary module, classified alarm data, response evaluation data and optimization data are integrated to generate an alarm analysis report, thereby providing a comprehensive decision-making basis for power plant alarm management.

[0162] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is therefore intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.

[0163] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A method for analyzing alarm data of a power plant, characterized in that: The following steps are involved: Step S1: collecting the original alarm signal; performing frequency calculation on the original alarm signal to obtain the frequency characteristics of the alarm signal; performing time series characteristic analysis on the frequency characteristics of the alarm signal to generate a time series alarm characteristic matrix, and performing visualization processing; Step S2: locating repeated alarm points on the time series alarm feature matrix to obtain repeated alarm points; clustering the alarm types of the repeated alarm points based on a preset evaluation time threshold to generate classified alarm data; Step S3: Based on the classified alarm data, inter-class correlation analysis is performed on the repeated alarm points to obtain alarm point correlation data; based on the alarm point correlation data, alarm optimization deduction is performed to generate deduced alarm optimization data; Step S4: Acquire alarm history records; perform alarm pattern recognition on the time series alarm feature matrix according to the alarm history records to obtain an alarm pattern sequence; perform flooding situation prediction on the alarm pattern sequence to generate a predicted alarm flooding situation; Step S5: performing alarm response evaluation on the classified alarm data based on the alarm history records to generate response evaluation data; pre-adjusting the deduced alarm optimization data based on the predicted alarm flooding situation and the response evaluation data to obtain pre-adjusted alarm optimization data; Step S6: Summarize the classified alarm data, response evaluation data and pre-adjustment alarm optimization data to generate an alarm analysis report.

2. The alarm data analysis method of a power plant according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: collecting the original alarm signal; performing digital filtering on the original alarm signal to obtain a filtered alarm signal; Step S12: segmenting the filtered alarm signal into time windows to obtain time period alarm data; performing overlapping window processing on the time period alarm data to obtain overlapping time period alarm data; Step S13: Perform frequency statistics processing on the alarm data of the overlapping period to generate the frequency characteristics of the alarm signal; perform time series correlation calculation on the frequency characteristics of the alarm signal to obtain time series characteristic data; Step S14: performing matrix conversion processing on the time series feature data to obtain a time series alarm feature matrix; and visually displaying the time series alarm feature matrix.

3. The alarm data analysis method for a power plant according to claim 1, characterized in that: Step S2 includes the following steps: Step S21: Perform frequency statistical analysis on the time series alarm feature matrix to obtain alarm frequency data; Step S22: performing threshold comparison and screening on the alarm frequency data according to a preset alarm frequency threshold to obtain repeated alarm data; locating repeated alarm points according to the repeated alarm data to obtain repeated alarm points; Step S23: performing time dimension aggregation on repeated alarm points to generate time clustering data; performing feature focusing on the time clustering data to obtain feature clustering data; Step S24: classify the feature clustering data by type labeling to generate classified alarm data.

4. The alarm data analysis method for a power plant according to claim 1, characterized in that: Step S3 includes the following steps: Step S31: constructing a feature vector for the classified alarm data to obtain feature description data; performing similarity calculation on the feature description data to generate a similarity matrix; Step S32: performing inter-class correlation analysis on repeated alarm points based on the similarity matrix to obtain alarm point correlation data; Step S33: performing optimization rule matching on the alarm point correlation data to obtain optimization strategy data; performing numerical simulation test based on the optimization strategy data to generate numerical simulation results; Step S34: Perform alarm optimization deduction on the optimization strategy data based on the numerical simulation results to generate deduced alarm optimization data.

5. The method for analyzing alarm data of a power plant according to claim 4, characterized in that: Step S32 includes the following steps: Based on the similarity matrix, the repeated alarm points are associated and mapped to obtain the alarm point dependency matrix; based on the alarm point dependency matrix, the repeated alarm points are associated and analyzed to generate the alarm point associated trajectory; Perform aggregation analysis based on alarm point correlation trajectories to obtain alarm point behavior patterns; extract inter-point behavior rules from alarm point behavior patterns to obtain inter-point behavior rules; The separation degree of repeated alarm points is calculated based on the behavior rules between points to generate the inter-class sparsity; A multi-layer correlation network is constructed according to the inter-class sparsity and alarm point dependency matrix to obtain the alarm point association network; the alarm point association network is analyzed layer by layer to generate alarm point correlation data.

6. The method for analyzing alarm data of a power plant according to claim 1, characterized in that: Step S4 includes the following steps: Step S41: Acquire alarm history records; perform time completion processing on the alarm history records to obtain completed history records; Step S42: performing periodic analysis on the time series alarm feature matrix based on the completed historical records to generate periodic feature data; performing alarm trend recognition on the periodic feature data to obtain trend feature data; Step S43: performing pattern decomposition on the trend feature data to obtain pattern component data; performing pattern matching on the pattern component data and the time series alarm feature matrix to generate an alarm pattern sequence; Step S44: Decompose the alarm pattern sequence to obtain pattern sequence features; perform time series change trend analysis on the pattern sequence features to generate alarm pattern change trends; Step S45: Perform situation assessment based on the alarm mode change trend to generate a predicted alarm flooding situation.

7. The method for analyzing alarm data of a power plant according to claim 1, characterized in that: Step S5 includes the following steps: Step S51: Analyze the response time of the alarm history records to obtain historical response time data; Step S52: mapping the classified alarm data to a response time according to the historical response time data to generate a predicted current alarm response time; Step S53: Use the preset index and alarm scoring formula to perform alarm response evaluation on the predicted current alarm response time to generate response evaluation data, wherein the alarm scoring formula is: Among them, S is the final score, V is the achieved value of the device, and T is the index of this value; Step S54: self-organizing mapping is performed on the predicted alarm flooding situation and response evaluation data to obtain mapping relationship data; and pre-adjusting the deduced alarm optimization data based on the mapping relationship data to obtain pre-adjusted alarm optimization data.

8. The method for analyzing alarm data of a power plant according to claim 7, characterized in that: Step S54 includes the following steps: Adaptively segment the predicted alarm flooding situation by thresholds to obtain flooding level data; decompose the response evaluation data into time segments to obtain time response characteristics; Cross-validate the flooding level data and time response characteristics to obtain the response weight matrix; Perform multi-dimensional decomposition on the deduction alarm optimization data to obtain the optimized feature vector; perform self-organizing mapping on the optimized feature vector according to the response weight matrix to obtain mapping relationship data; Compensation condition analysis is performed based on flooding level data and time response characteristics to generate compensation condition parameters; compensation characteristic conversion is performed on the compensation condition parameters based on mapping relationship data to generate compensation characteristic vectors; The deduced alarm optimization data is pre-adjusted by compensating the characteristic vector to obtain pre-adjusted alarm optimization data.

9. The method for analyzing alarm data of a power plant according to claim 1, characterized in that: Step S6 includes the following steps: Step S61: Perform multi-dimensional statistics on the classified alarm data to obtain statistical dimension data; perform index calculation on the statistical dimension data to obtain classified statistical data; Step S62: convert the classified statistical data into a report form to generate an alarm classification statistical report; Step S63: Calculate the performance index of the response evaluation data to obtain performance index data; perform benchmark analysis on the performance index data to generate performance benchmark data; Step S64: convert the performance benchmarking data into a report form to generate a response performance report; Step S65: Based on a preset report template, the alarm classification statistics report, the response performance report and the pre-adjusted alarm optimization data are integrated to generate an alarm analysis report.

10. An alarm data analysis system for a power plant, characterized in that: The method for analyzing the alarm data of a power plant according to claim 1 is used to perform the alarm data analysis system of the power plant, the alarm data analysis system of the power plant comprising: The data visualization module is used to collect the original alarm signal; perform frequency calculation on the original alarm signal to obtain the frequency characteristics of the alarm signal; perform time series characteristic analysis on the frequency characteristics of the alarm signal to generate a time series alarm characteristic matrix and perform visualization processing; The repeated alarm analysis module is used to locate the repeated alarm points in the time series alarm feature matrix to obtain the repeated alarm points; cluster the alarm types of the repeated alarm points based on the preset evaluation time threshold to generate classified alarm data; The alarm correlation analysis module is used to perform inter-class correlation analysis on repeated alarm points based on classified alarm data to obtain alarm point correlation data; perform alarm optimization deduction based on the alarm point correlation data to generate deduced alarm optimization data; The alarm flood prediction module is used to obtain alarm history records; perform alarm pattern recognition on the time series alarm feature matrix according to the alarm history records to obtain the alarm pattern sequence; perform flood situation prediction on the alarm pattern sequence to generate the predicted alarm flood situation; The alarm response analysis module is used to evaluate the alarm response of the classified alarm data based on the alarm history records and generate response evaluation data; pre-adjust the deduced alarm optimization data based on the predicted alarm flooding situation and response evaluation data to obtain pre-adjusted alarm optimization data; The report summary module is used to summarize the classified alarm data, response evaluation data and pre-adjustment alarm optimization data to generate an alarm analysis report.

Images