Classification analysis method and system between micro working conditions and micro intervals of complex system

Abstract

The invention relates to a classification analysis method and system between micro working conditions and micro intervals of a complex system. The method comprises the steps of obtaining and preprocessing operation data of the complex system; according to the operation principle and performance influence factors of the complex system, key features capable of representing the operation state of the system are extracted from the preprocessed data; data points corresponding to the key features are divided into different clustering clusters through clustering, and each cluster represents a micro working condition / micro interval; according to a clustering result, determining a category label for each micro-working condition / micro-interval, determining system operation characteristics and system performance indexes in the micro-working condition / micro-interval by establishing a micro-working condition / micro-interval classification model and a corresponding performance analysis model, establishing a correlation model of each micro-working condition performance index and a micro-working condition performance parameter, and determining the micro-working condition / micro-interval. And determining an optimal index through optimization for a subsequent optimization process. The classification of the operation states of the complex system is realized, and powerful support is provided for optimized operation.

Description

Technical Field

[0001] The present invention relates to the technical field of data analysis, and in particular to a classification analysis method and system for micro-conditions / micro-intervals of complex systems. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] In the industrial field, the processes of product manufacturing, energy production, and equipment operation services are complex, such as energy and electricity, chemical production, and steel manufacturing. The various data generated during the operation of such complex systems can be used to optimize the operating status. However, these complex systems often have multivariable, strong coupling, nonlinear, and time-varying characteristics. Their operation process is affected by many factors, and the operating conditions change frequently. Traditional analysis methods are difficult to accurately describe the operating status of the system under different working conditions, and cannot meet the needs of refined operation and optimization of the system.

[0004] For example, in coal-fired power units, changes in factors such as unit load, coal type characteristics, and environmental conditions can lead to significant differences in the system's operating status. Existing technologies are unable to effectively process operating data under such complex conditions, resulting in an inability to accurately grasp system performance and difficulty in achieving efficient and economical operation. Summary of the invention

[0005] In order to solve the technical problems existing in the above-mentioned background technology, the present invention provides a classification and analysis method and system for micro-conditions / micro-intervals of complex systems, which solves the problem that the existing technology is difficult to cope with the changeable working conditions of complex systems and difficult to achieve precise operation optimization, and realizes accurate classification and in-depth analysis of the operating status of complex systems, providing strong support for the optimized operation of complex systems.

[0006] In order to achieve the above object, the present invention adopts the following technical solution:

[0007] A first aspect of the present invention provides a classification and analysis method for micro-conditions / micro-intervals of a complex system, comprising the following steps:

[0008] Data collection and processing: obtain and pre-process the operating data of complex systems;

[0009] Working condition cluster analysis: According to the operating principle and performance influencing factors of complex systems, key features that can characterize the operating status of the system are extracted from the preprocessed data, and the data points corresponding to the key features are divided into different clusters, each of which represents a micro-working condition / micro-interval;

[0010] Analyze the performance indicators of each micro-condition and establish a correlation model between the performance indicators and the operating parameters of the micro-conditions: According to the clustering results, determine the category label for each micro-condition / micro-interval, and determine the system operating characteristics and system performance indicators within the micro-condition / micro-interval by establishing a micro-condition / micro-interval classification model and a corresponding performance analysis model, and establish a correlation model between each micro-condition performance indicator and the micro-condition performance parameters;

[0011] Index optimization: According to the system operating characteristics within micro-operating conditions / micro-intervals, the distribution law and correlation of operating parameters and the degree of influence on system performance indicators are determined and saved;

[0012] Guiding operation and operation optimization: Determine the gap between the actual operating performance index of each micro-operating condition and the optimal index value, and determine the optimization technical route and method based on the correlation model of the optimal index value, and determine the adjustment method of related parameters; through the optimization of the indicators of each micro-operating condition, the optimization of all operating indicators of the complex system can be achieved.

[0013] Furthermore, the operation data of the complex system is obtained, specifically: a sensor network is deployed at a set location of the complex system to obtain the operation data, and the operation data at least includes the equipment operation parameters of the complex system, environmental parameters, energy parameters input into the complex system, and performance indicators of the output products.

[0014] Furthermore, preprocessing is specifically: processing the acquired operation data using at least one of data cleaning, interpolation, filtering and normalization.

[0015] Furthermore, features that can characterize the operating status of the system are extracted from the preprocessed data, and features whose contribution rate to the operating status is greater than a set value are selected as key features.

[0016] Furthermore, features that can characterize the operating status of the system are extracted, and based on the principal component analysis method, the correlation analysis method or the kernel principal component analysis method, features whose contribution rate to the operating status is greater than a set value are screened out as key features.

[0017] Furthermore, through clustering, the data points corresponding to the key features are divided into different clustering clusters. Specifically, a density-based clustering algorithm, a hierarchical clustering algorithm or a fuzzy clustering algorithm is used to perform clustering analysis on the operating conditions of the complex system according to the selected key features. By setting the neighborhood radius and the minimum number of sample points, the data during the operation of the complex system is divided into different clustering clusters. Each cluster represents a micro-condition / micro-interval, reflecting different operating condition intervals of the system operation process.

[0018] Furthermore, the clustering results are evaluated and the parameters in the clustering algorithm are adjusted according to the evaluation results.

[0019] Furthermore, complex systems include energy and power systems, chemical production systems, steel manufacturing systems, wind power generation systems or urban rail transit systems.

[0020] Furthermore, a mathematical model between system performance indicators and operating parameters is established using multivariate linear regression or neural network methods to determine the system operating characteristics and system performance indicators within micro-conditions / micro-intervals, and to establish an association model between each micro-condition performance indicator and micro-condition performance parameter.

[0021] A second aspect of the present invention provides a classification and analysis system for micro-conditions / micro-intervals of complex systems, comprising:

[0022] The data acquisition and preprocessing module is configured to: acquire and preprocess the operating data of the complex system;

[0023] The feature extraction and selection module is configured to: extract key features that can characterize the system operation status from the preprocessed data according to the operation principle and performance influencing factors of the complex system;

[0024] The micro-condition / micro-interval clustering analysis module is configured to: divide the data points corresponding to the key features into different clusters through clustering, and each cluster represents a micro-condition / micro-interval;

[0025] The micro-condition / micro-interval classification module is configured to: determine a category label for each micro-condition / micro-interval according to the clustering result, determine the system operation characteristics and system performance indicators within the micro-condition / micro-interval by establishing a micro-condition / micro-interval classification model and a corresponding performance analysis model, and establish a correlation model between each micro-condition performance indicator and the micro-condition performance parameter;

[0026] The micro-condition / micro-interval analysis application module is configured to determine the distribution law and correlation of operating parameters and the degree of influence on system performance indicators according to the system operating characteristics within the micro-condition / micro-interval.

[0027] Compared with the prior art, one or more of the above technical solutions have the following beneficial effects:

[0028] 1. By analyzing the operating data of complex systems in industrial production (such as energy and power systems, chemical production systems, steel manufacturing systems, wind power generation systems or urban rail transit systems), the key features that can characterize the operating status of the system are selected, and the data is processed by clustering, so that the operating status data is divided into data clusters representing multiple micro-conditions / micro-intervals, and the complex product manufacturing processes and service processes in the industrial field are reasonably divided into multiple micro-conditions / micro-intervals, making the description of the system operating status more refined, and providing the possibility for subsequent targeted operation optimization.

[0029] 2. By characterizing data clusters between multiple micro-conditions / micro-intervals and building classification models based on data cluster classification labels, we can deeply understand the operating characteristics and laws of complex systems, provide detailed decision-making basis for subsequent optimization of system operation, and help achieve goals such as improving system efficiency, reducing energy consumption, and extending equipment life. It has strong versatility and adaptability and can be applied to a variety of complex systems, such as energy and power systems, chemical production systems, industrial manufacturing systems, etc., providing effective technical support for the operation optimization of complex systems in different fields.

[0030] 3. Use historical operating data to determine the optimal value of each micro-condition performance indicator as the optimal standard for the indicator of the micro-condition. Through the gap between the actual operating condition performance indicator of each micro-condition and the optimal indicator value, combined with the correlation model of the optimal indicator value, determine the optimization technical route and method. By optimizing each micro-condition indicator, the optimization of all operating indicators of the complex system can be achieved.

[0031] 4. Through comprehensive data collection and preprocessing, accurate information during the operation of complex systems can be obtained, providing a reliable data basis for subsequent analysis and reducing the impact of data errors on analysis results.

[0032] 5. The process of feature extraction and selection effectively reduces the data dimension, highlights key information, improves analysis efficiency and accuracy, and helps focus on factors that have a greater impact on the system's operating status. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0034] Figure 1 It is a schematic diagram of a classification analysis process of micro-conditions / micro-intervals of a complex system provided by one or more embodiments of the present invention;

[0035] Figure 2 It is a schematic diagram of a classification analysis process of complex system micro-conditions / micro-intervals provided by one or more embodiments of the present invention. DETAILED DESCRIPTION

[0036] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0037] It should be noted that the following detailed descriptions are exemplary and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0038] It should be noted that the terms herein are only intended to describe specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0039] Industrial fields, such as energy and electricity, chemical production, and steel manufacturing, cover multiple links such as product research and development, production, operation and maintenance, management, and services, and have a wide variety of data types, including structured data and unstructured data, such as production data, equipment status data, and behavioral data. Secondly, due to the variety of machine types that support production within industrial enterprises, data comes directly from various production equipment, so the data formats are also different. In addition, industrial data may be subject to various interferences during the collection, transmission, and storage processes, resulting in uneven data quality and increasing the difficulty of data processing. Therefore, it is called a complex system.

[0040] By analyzing the data of complex systems, we can more accurately understand the resource consumption in the production process, so as to make reasonable resource allocation, avoid resource waste and improve production efficiency. By optimizing the data of complex systems, we can timely discover potential failure risks, reduce equipment failures and downtime through predictive maintenance, and further reduce production costs.

[0041] Traditional analysis methods are difficult to accurately describe the operating status of complex systems under different working conditions, and cannot meet the needs of refined operation and optimization of the system. Therefore, the following embodiment provides a classification analysis method and system for micro-conditions / micro-intervals of complex systems, which can accurately identify different operating conditions of complex systems, reasonably divide micro-conditions / micro-intervals, and make the description of the system operating status more refined. By deeply understanding the operating characteristics and laws of the system within each micro-condition / micro-interval, a detailed decision-making basis is provided for the optimized operation of the system, which helps to achieve the goals of improving system efficiency, reducing energy consumption, and extending equipment life.

[0042] Embodiment 1:

[0043] like Figure 1-Figure 2 As shown, a classification analysis method for micro-conditions / micro-intervals of complex systems includes the following steps:

[0044] Data collection and preprocessing, specifically: obtaining and preprocessing the operating data of complex systems;

[0045] Feature extraction and selection, specifically: extract key features that can characterize the system's operating status from preprocessed data based on the operating principles and performance influencing factors of the complex system;

[0046] Cluster analysis of micro-conditions / micro-intervals: specifically, by clustering, the data points corresponding to the key features are divided into different clusters, each cluster representing a micro-condition / micro-interval;

[0047] Classification and modeling of micro-conditions / micro-intervals, specifically: according to the clustering results, determine the category label for each micro-condition / micro-interval, establish a micro-condition / micro-interval classification model and a corresponding performance analysis model, determine the operating characteristics and system performance indicators of the system within the micro-condition / micro-interval, and establish a correlation model between each micro-condition performance indicator and the micro-condition performance parameter;

[0048] Indicator optimization, specifically: determine the optimal value of each micro-operating condition performance indicator based on historical operating data, and use it as the optimal indicator standard for the micro-operating condition;

[0049] Analysis and application, specifically: determine the gap between the actual operating performance index of each micro-operating condition and the optimal index value, and determine the optimization technical route and method based on the correlation model of the optimal index value, and determine the adjustment method of related parameters; through the optimization of the indicators of each micro-operating condition, the optimization of all operating indicators of the complex system is achieved.

[0050] Data collection and preprocessing, specifically:

[0051] Deploy sensor networks at key locations of complex systems to comprehensively collect system operation data, including but not limited to equipment operation parameters (such as temperature, pressure, flow, speed, etc.), environmental parameters (such as temperature, humidity, air pressure, etc.), input energy parameters (such as coal composition, fuel calorific value, etc.), and system output performance indicators (such as power, efficiency, product quality, etc.);

[0052] Clean the collected data, remove outliers and erroneous data, and supplement missing data through data interpolation, filtering and other methods to ensure the accuracy and completeness of the data;

[0053] The preprocessed data is normalized and the data of different physical dimensions are mapped to a specific interval for subsequent calculation and analysis.

[0054] Feature extraction and selection, specifically:

[0055] According to the system operation principle and performance influencing factors, determine the key features that can characterize the system operation status. For example, for coal-fired power units, select the turbine steam inlet pressure, temperature, flow, boiler exhaust temperature, oxygen content, feedwater pump power consumption rate, etc. as features;

[0056] The principal component analysis (PCA) and correlation analysis methods are used to screen out the main features that contribute most to the distinction of operating status, reduce the data dimension and improve the analysis efficiency.

[0057] Cluster analysis between micro-conditions and micro-intervals, specifically:

[0058] Use density-based clustering algorithms (such as DBSCAN), hierarchical clustering algorithms or fuzzy clustering algorithms to cluster the system operating conditions according to the selected features. Taking the DBSCAN algorithm as an example, set the appropriate neighborhood radius and minimum number of sample points to divide the data points into different clusters, each cluster represents a micro-condition / micro-interval;

[0059] The clustering results can be evaluated and optimized. Evaluation indicators such as silhouette coefficient and Calinski-Harabasz index can be used to measure the clustering effect. Clustering parameters or algorithms can be adjusted according to the evaluation results to ensure the rationality and effectiveness of the clustering results.

[0060] Micro-condition / micro-interval classification and modeling, specifically:

[0061] According to the clustering results, determine the category label for each micro-condition / micro-interval and establish a micro-condition / micro-interval classification model. The classification model can be constructed using algorithms such as decision tree, support vector machine (SVM) or neural network;

[0062] For each micro-operating condition / micro-interval, a corresponding performance analysis model is established. For example, a mathematical model between system performance indicators (such as efficiency, energy consumption, etc.) and operating parameters is established using methods such as multivariate linear regression and neural networks, so as to deeply analyze the operating characteristics of the system within each micro-operating condition / micro-interval.

[0063] Analysis and application, specifically:

[0064] Determine the gap between the actual operating performance index and the optimal index value of each micro-operating condition, and determine the optimization technology route and method based on the correlation model of the optimal index value, and determine the adjustment method of related parameters; through the optimization of the index of each micro-operating condition, the optimization of all operating indicators of the complex system is realized. By analyzing the distribution law, correlation and influence of the system operating parameters on the performance indicators in each micro-operating condition / micro-interval. For example, in a certain micro-operating condition interval of a coal-fired power unit, the relationship between the turbine efficiency and the steam inlet parameters is studied to determine the key influencing factors.

[0065] Based on the analysis results, decision support is provided for the system to optimize operation, such as formulating equipment operation strategies under different micro-operating conditions / micro-intervals (such as adjusting equipment operating parameters, optimizing equipment start-stop sequence, etc.), maintenance plans (such as determining equipment maintenance cycles and key maintenance areas based on operating conditions), and energy management plans (such as reasonably selecting energy input methods and parameters) to achieve efficient and economical operation of the system.

[0066] The time interval for data collection is determined according to the dynamic characteristics of the system. For system parameters that change rapidly, a shorter collection interval is used to ensure that changes in the system operating status can be accurately captured.

[0067] In the feature extraction and selection steps, for complex systems with strong nonlinear relationships, nonlinear feature extraction methods such as kernel principal component analysis (KPCA) are used to better mine the nonlinear feature information in the data and improve the accuracy of classification and analysis.

[0068] In the micro-condition / micro-interval clustering analysis step, when the fuzzy clustering algorithm is used, the fuzzy membership function is introduced to describe the degree of membership of the data points to each clustering cluster, so as to more flexibly handle the fuzzy boundary of the condition in the complex system and improve the rationality of clustering.

[0069] In the micro-condition / micro-interval classification and modeling steps, for multi-classification problems, an integrated learning method (such as random forest) is used to build a classification model to improve the generalization and accuracy of the classification model and enhance the classification ability of different working conditions of complex systems.

[0070] In the micro-condition / micro-interval analysis and application steps, real-time monitoring data is combined with the use of predictive models to predict the future performance of the system, provide early warning of possible performance degradation or failure, and provide forward-looking guidance for system operation adjustment and maintenance.

[0071] By analyzing the operating data of complex systems in industrial production (such as energy and power systems, chemical production systems, steel manufacturing systems, wind power generation systems or urban rail transit systems), key features that can characterize the operating status of the system are selected, and the data is processed by clustering, so that the operating status data is divided into data clusters that characterize multiple micro-conditions / micro-intervals. The complex product manufacturing processes and service processes in the industrial field are reasonably divided into multiple micro-conditions / micro-intervals, making the description of the system operating status more refined, and providing the possibility for subsequent targeted operation optimization.

[0072] The historical operating data is used to determine the optimal value of each micro-condition performance indicator as the optimal standard for the indicator of the micro-condition. The gap between the actual operating condition performance indicator of each micro-condition and the optimal indicator value is combined with the correlation model of the optimal indicator value to determine the optimization technical route and method. By optimizing each micro-condition indicator, the optimization of all operating indicators of the complex system is achieved.

[0073] By characterizing data clusters between multiple micro-conditions / micro-ranges and building classification models based on data cluster classification labels, we can gain an in-depth understanding of the operating characteristics and laws of complex systems, provide detailed decision-making basis for subsequent optimization of system operation, and help achieve goals such as improving system efficiency, reducing energy consumption, and extending equipment life. The method has strong versatility and adaptability and can be applied to a variety of complex systems, such as energy and power systems, chemical production systems, industrial manufacturing systems, etc., providing effective technical support for the operation optimization of complex systems in different fields.

[0074] Through comprehensive data collection and preprocessing, the above process can obtain accurate information during the operation of complex systems, provide a reliable data basis for subsequent analysis, and reduce the impact of data errors on analysis results.

[0075] The feature extraction and selection steps in the above process effectively reduce the data dimension, highlight key information, improve analysis efficiency and accuracy, and help focus on factors that have a greater impact on the system operating status.

[0076] By performing cluster analysis on the collected data and using the obtained data clusters to accurately identify the different operating conditions of complex systems, the complex product manufacturing process in the industrial field can be reasonably divided into multiple micro-conditions / micro-intervals, making the description of the system operating status more refined and providing the possibility for targeted operation optimization.

[0077] The classification model constructed by using the classification labels of micro-conditions / micro-intervals can deeply understand the operating characteristics and laws of the system, provide detailed decision-making basis for system optimization operation, and help achieve goals such as improving system efficiency, reducing energy consumption, and extending equipment life.

[0078] It has strong versatility and adaptability, and can be applied to a variety of complex systems, such as energy and power systems, chemical production systems, industrial manufacturing systems, etc., providing effective technical support for the operation optimization of complex systems in different fields.

[0079] Figure 1 It is a micro-condition / micro-interval clustering analysis diagram, which gives the key operating conditions 1, 2...i of the complex system determined through analysis. There are n, m...k operating conditions under each operating condition. It is a reference diagram for clustering analysis of the unit's micro-condition / micro-interval using i key operating conditions and n, m...k operating conditions under each operating condition.

[0080] This embodiment takes a coal-fired power unit as an example. High-precision sensors are deployed in key equipment such as the unit's turbine, boiler, generator, feed water pump, and environmental monitoring points. Sensor types include temperature sensors, pressure sensors, flow sensors, component analyzers, etc. The collection time interval is set to 1 minute (for parameters that change rapidly, such as steam pressure and temperature) to 10 minutes (for parameters that change slowly, such as equipment wear).

[0081] The collected data include unit load (0-100% rated load), ambient temperature (-20℃-40℃), humidity (20%-90%), air pressure (95kPa-105kPa), calorific value of coal type (18MJ / kg-25MJ / kg), volatile matter (10%-30%), ash content (20%-40%), steam inlet pressure at each level of turbine (0.1MP a-5MPa) and temperature (50℃-400℃), exhaust pressure (3kPa-10kPa), extraction pressure of each level (0.1MPa-2MPa) and temperature (50℃-300℃), boiler main steam flow (100t / h-1000t / h), feed water temperature (100℃-250℃), exhaust temperature (100℃-200℃), exhaust oxygen content (3%-8%), feed water pump speed (1000rpm-5000rpm), power consumption rate (0.1%-1%), generator output power (0-1000MW), etc.

[0082] After the data is collected, it is first checked, and the obviously erroneous data (such as abnormally high or low temperature data caused by temperature sensor failure, data with sudden jumps in pressure sensor readings, etc.) is deleted or corrected, and the missing data is supplemented by linear interpolation. Then the data is normalized, for example, the pressure data is normalized to the [0,1] interval, the temperature data is normalized to the [-1,1] interval, etc., so that the data of different physical dimensions are comparable.

[0083] In the feature extraction and selection stage, according to the operating principle and performance influencing factors of coal-fired power units, steam turbine inlet pressure, temperature, exhaust pressure, extraction pressure and temperature, boiler exhaust temperature, feed water temperature, feed water pump power consumption rate, coal calorific value, etc. are selected as key features. Due to the complex nonlinear relationship between various parameters in the operation of coal-fired power units, the kernel principal component analysis (KPCA) method is used for feature extraction, and the principal components with a cumulative contribution rate of more than 85% are selected as the main features, reducing the data dimension from more than ten original ones to 3-5, reducing the subsequent calculation amount while retaining key information.

[0084] For cluster analysis of micro-conditions / micro-intervals, the density-based DBSCAN algorithm is used. According to the characteristics and experience of coal-fired power unit operation data, the neighborhood radius is set to 0.5 (normalized data space) and the minimum number of sample points is 5. The collected data after preprocessing and feature selection is input into the DBSCAN algorithm for cluster analysis. The clustering results obtained 5 main micro-conditions / micro-intervals, corresponding to different unit load ranges and operating states, such as high-load stable operation area, medium-load fluctuation area, low-load operation area, startup process area and shutdown process area. The clustering results were evaluated using the silhouette coefficient, and it was found that the clustering effect was good, the data points within each micro-condition / micro-interval were relatively compact, and the distinction between different intervals was obvious.

[0085] In terms of classification and modeling of micro-conditions / micro-intervals, support vector machines (SVM) are used to build classification models to distinguish different micro-conditions / micro-intervals. The SVM model is trained using historical operating data, and appropriate kernel functions (such as radial basis functions RBF) and model parameters (such as penalty factor C and kernel function parameter γ) are selected to optimize model performance through cross-validation and other methods. For each micro-condition / micro-interval, a performance analysis model is established using the multivariate linear regression method. For example, in the high-load stable operation area, a linear regression model is established between the unit's power generation efficiency and the turbine's steam inlet parameters, exhaust pressure, boiler exhaust temperature, etc.; in the low-load operation area, a model is established between the feedwater pump power consumption rate and the unit load, feedwater temperature, etc. These models can be used to deeply analyze the quantitative relationship between system operating parameters and performance indicators in each micro-condition / micro-interval.

[0086] In the micro-condition / micro-interval analysis and application stage, the distribution law and correlation of system operating parameters in each micro-condition / micro-interval are analyzed. It is found that in the high-load stable operation area, the turbine inlet steam pressure and temperature are positively correlated with the power generation efficiency, and the correlation is strong; while in the low-load operation area, the feedwater pump power consumption rate is negatively correlated with the unit load, that is, the lower the load, the higher the feedwater pump power consumption rate.

[0087] Based on these analysis results, different operation strategies are formulated for coal-fired power units. When operating at high load and stable conditions, the steam inlet parameters of the steam turbine are optimized to keep it in the high efficiency range, while paying close attention to the exhaust temperature of the boiler to prevent excessive exhaust heat loss; when operating at low load, the operation mode of the feedwater pump is reasonably adjusted, such as using variable frequency speed regulation technology to reduce the speed of the feedwater pump and reduce the power consumption rate. In addition, combined with real-time monitoring data, the established performance analysis model is used to predict the future performance of the unit.

[0088] For example, when it is predicted that the power generation efficiency of the unit will decrease within a certain micro-operating condition / micro-interval, check the operating status of related equipment in advance, such as the scaling of turbine blades and the cleanliness of boiler heating surfaces, and take timely measures, such as turbine cleaning and boiler soot blowing, to prevent further performance degradation and ensure efficient and economical operation of the unit.

[0089] The above embodiments are only specific explanations of the present invention. In practical applications, data collection, processing, analysis methods, model building and other aspects can be adjusted and optimized accordingly according to the characteristics and requirements of different complex systems to achieve the best operation optimization effect.

[0090] Example 2: Chemical production process optimization

[0091] Data collection and preprocessing, specifically:

[0092] In chemical production equipment, various sensors are arranged in key equipment such as reactors, distillation towers, heat exchangers, pumps, as well as raw material supply pipelines and environmental monitoring points. Sensors include temperature sensors (measuring reaction temperature range of -20℃-250℃), pressure sensors (measuring range of 0.1MPa-10MPa), flow sensors (measuring raw material and product flow range of 0.1m3 / h-100m3 / h), concentration sensors (detecting reactant and product concentration range of 0%-100%), etc. The collection time interval is set according to the dynamic characteristics of the process. For parameters related to fast reaction processes, such as reactor temperature and pressure, data are collected every 10 seconds; for relatively slowly changing parameters, such as equipment corrosion rate monitoring data, data are collected every 1 hour.

[0093] The collected data is transmitted to the data processing center. The data cleaning module first checks the data integrity and accuracy, removes obviously erroneous data, such as abnormal values ​​caused by sensor failure (such as pressure suddenly becoming negative or extremely large), and uses the mean filling method (applicable to data with small fluctuations) or the model-based prediction filling method (applicable to data with certain regularities) to process missing data. Then the data is normalized and all data is mapped to the [-1,1] interval for subsequent calculations.

[0094] Feature extraction and selection, specifically:

[0095] Determine key features based on chemical production principles and factors affecting product quality and output. For example, for a chemical reaction process, select reactor temperature, pressure, raw material concentration, catalyst activity, reaction time, distillation tower top and bottom temperatures, heat exchanger heat transfer efficiency, etc. as features.

[0096] Due to the complex nonlinear relationship in the chemical process, Kernel principal component analysis (KPCA) is used for feature extraction. After KPCA processing, the principal components with a cumulative contribution rate of more than 90% are selected as the main features, reducing the original dozen feature dimensions to 4-6, effectively reducing data redundancy while retaining key information.

[0097] Cluster analysis between micro-conditions and micro-intervals, specifically:

[0098] The fuzzy clustering algorithm (FCM) is used to cluster the operating conditions of the chemical production process. Based on experience and data analysis results, the fuzzy index m=2 is set, and the number of cluster centers is 4 (preliminary division according to different production stages and product quality levels). The chemical production process is divided into different micro-conditions / micro-intervals through the FCM algorithm, such as the raw material preheating stage, the initial reaction stage, the peak reaction period, the product separation stage and other different operating condition intervals.

[0099] The clustering results were evaluated using the Calinski-Harabasz index, and it was found that the clustering effect was good, the data points within each micro-condition / micro-interval had high similarity, and there was a clear distinction between different intervals. For situations where the evaluation index is not ideal, the fuzzy index m or the number of cluster centers is appropriately adjusted, and the clustering analysis is repeated until a satisfactory clustering result is obtained.

[0100] Micro-condition / micro-interval classification and modeling, specifically:

[0101] The decision tree algorithm is used to build a micro-condition / micro-interval classification model. The collected data after preprocessing and feature selection is used as input, and the micro-condition / micro-interval category label is used as output to train the decision tree model. Pruning and other technologies are used to prevent model overfitting and improve the generalization ability of the model.

[0102] For each micro-condition / micro-interval, a neural network is used to establish a performance analysis model between product quality and output and operating parameters. For example, in the micro-condition interval during the peak reaction period, a neural network model is established between product output and reactor temperature, pressure, raw material concentration, and catalyst activity; in the product separation stage, a model is established between product purity and the temperature and pressure at the top and bottom of the distillation tower. The neural network model is trained with a large amount of historical data, and the network structure and parameters are adjusted so that the model can accurately predict product quality and output under different conditions.

[0103] Micro-condition / micro-interval analysis and application, specifically:

[0104] The distribution and correlation of operating parameters in each micro-condition / micro-interval were analyzed. It was found that during the peak reaction period, the reactor temperature was positively correlated with the product output, but when the temperature exceeded a certain threshold, the product quality would decrease; the raw material concentration was positively correlated with the output within a certain range, but too high or too low would affect the reaction efficiency.

[0105] Develop an optimization strategy based on the analysis results. During the peak reaction period, the reactor temperature is precisely controlled by adjusting the cooling medium flow rate to keep it within the optimal reaction temperature range. At the same time, the feed rate of raw materials is adjusted according to the real-time monitoring results of the raw material concentration to ensure efficient reaction and qualified product quality. In the product separation stage, according to the product purity prediction model, the operating parameters of the distillation tower, such as reflux ratio, plate temperature, etc., are optimized to improve product separation efficiency and purity. In addition, combined with real-time monitoring data, the performance analysis model is used to predict product quality and output. When it is predicted that product quality may decline or output may decrease, an early warning is issued in time, and production parameters are adjusted in advance or the equipment operation status is checked, such as whether the catalyst is deactivated or whether the heat exchanger is fouled, etc., to ensure the stable and efficient operation of the chemical production process and improve product quality and production efficiency.

[0106] Example 3: Steel Manufacturing Process Optimization

[0107] Data collection and preprocessing, specifically:

[0108] Sensors are installed in key parts such as blast furnaces, converters, rolling mills, and related heating furnaces and cooling equipment in the main processes of ironmaking, steelmaking, and steel rolling in steel manufacturing. Sensor types include temperature sensors (measuring range from room temperature to above 1600°C, used to monitor molten iron temperature, billet heating temperature, etc.), pressure sensors (measuring range 0.1MPa-5MPa, used to detect blast furnace pressure, steam pressure, etc.), flow sensors (measuring range 0.1m3 / s-10m3 / s, used to measure gas flow, cooling water flow, etc.), and stress strain sensors (used to measure rolling force of rolling mills, equipment structural stress, etc.). The frequency of data collection is determined according to the characteristics of the process. For example, the key parameters of the ironmaking blast furnace are collected every 5 minutes, the important parameters in the steelmaking converter blowing process are collected every 1 minute, and the parameters such as rolling force in the steel rolling process are collected in real time (once every 0.1 seconds).

[0109] After the collected data is transmitted to the data processing center, the data cleaning submodule first cleans the data. For abnormal data, such as erroneous data caused by strong electromagnetic interference of the sensor, it is deleted. For missing data, if the data missing rate is low (less than 5%), the average value of adjacent data is used to fill it; if the missing rate is high, it is filled with the predicted value based on the time series model according to the change law of the parameter in different production stages. Then the data is normalized, the temperature data is normalized to the [0,1] interval, the pressure data is normalized to the [-1,1] interval, etc., so that the data can be analyzed in the same dimension for subsequent analysis.

[0110] Feature extraction and selection, specifically:

[0111] According to the factors affecting the steel manufacturing process and product performance, key features are selected. In the ironmaking process, the blast furnace charge ratio (the ratio of raw materials such as ore and coke), blast temperature, blast humidity, furnace top pressure, etc. are selected as features; in the steelmaking process, the molten iron composition (carbon, silicon, manganese, etc.), converter oxygen blowing amount, slag material addition amount, etc. are selected; in the steel rolling process, the billet heating temperature, rolling speed, roll gap, etc. are selected.

[0112] Considering the complex nonlinear relationship between various parameters in the steel manufacturing process, kernel principal component analysis (KPCA) is used for feature extraction. After KPCA processing, the principal components with a cumulative contribution rate of more than 88% are selected as the main features, reducing the original multiple feature dimensions to 3-5, highlighting key information and reducing the amount of data processing.

[0113] Cluster analysis between micro-conditions and micro-intervals, specifically:

[0114] A hierarchical clustering algorithm is used to cluster the operating conditions of the steel manufacturing process. According to the production process and equipment operation characteristics, the termination condition of clustering is set as the distance between clusters is greater than a certain threshold (determined by analyzing historical data and process requirements) or the number of clusters reaches a predetermined value (based on the main production stages and product types, such as normal smelting of ironmaking, abnormal furnace conditions, early blowing, mid-blowing, late blowing of steelmaking, rough rolling, finishing rolling of steel rolling, etc. The number of clusters is initially set in different stages). The steel manufacturing process is divided into different micro-conditions / micro-intervals through a hierarchical clustering algorithm.

[0115] Use the silhouette coefficient to evaluate the clustering results. If the silhouette coefficient is low, it means that the clustering effect is not ideal. At this time, adjust the clustering distance measurement method (such as changing from Euclidean distance to Mahalanobis distance) or the clustering termination condition, and re-perform the clustering analysis until a reasonable clustering result is obtained.

[0116] Micro-condition / micro-interval classification and modeling, specifically:

[0117] A support vector machine (SVM) is used to build a micro-condition / micro-interval classification model, with the collected preprocessed and feature-selected data as input and the micro-condition / micro-interval category label as output to train the SVM model. The optimal kernel function (such as linear kernel function or radial basis function) and model parameters (such as penalty factor C and kernel function parameter γ) are selected through cross-validation and grid search methods to improve the accuracy and generalization ability of the classification model.

[0118] For each micro-condition / micro-interval, a performance analysis model between product quality (such as steel strength, hardness, toughness, etc.) and production efficiency (such as output per unit time) and operating parameters is established using the multivariate nonlinear regression method. For example, in the mid-term micro-condition interval of steelmaking blowing, a nonlinear regression model is established between the carbon content of molten steel and the amount of oxygen blowing, the amount of slag material added, and the initial carbon content of molten iron; in the micro-condition interval of steel rolling finishing, a model is established between steel strength and the gap between the rolls, the rolling speed, and the heating temperature of the billet. The model is trained and optimized through a large amount of historical production data, so that the model can accurately reflect the relationship between production parameters and product quality and production efficiency in each micro-condition / micro-interval.

[0119] Micro-condition / micro-interval analysis and application, specifically:

[0120] The distribution law and correlation of operating parameters in each micro-condition / micro-interval are deeply analyzed. In the normal smelting micro-condition interval of the iron-making blast furnace, it is found that the blast temperature is positively correlated with the furnace top pressure, and the blast temperature has an important influence on the molten iron output and quality; in the micro-condition interval of steel finishing rolling, the rolling speed is negatively correlated with the steel strength, and the reasonable adjustment of the roll gap is crucial to the dimensional accuracy and surface quality of the steel.

[0121] Develop optimization strategies based on the analysis results. During normal smelting in the iron-making blast furnace, the blast temperature and blast humidity are accurately controlled according to the furnace condition monitoring data and performance analysis model, the raw material ratio is optimized, and the molten iron output and quality are improved; during the finishing rolling process, the roll gap and rolling speed are dynamically adjusted according to the steel performance requirements and real-time monitoring data to ensure that the quality of steel products meets the standard requirements and improve production efficiency. In addition, combined with real-time monitoring data and performance analysis models, product quality and production efficiency are predicted. When it is predicted that the strength of the steel may not meet the standard or the production efficiency will decrease, the production parameters are adjusted in time, such as increasing or decreasing the roll reduction, adjusting the heating furnace temperature, etc., and the equipment is inspected and maintained, such as checking the wear of the rolls and the working status of the heating furnace burners, to ensure the stable and efficient operation of the steel manufacturing process, reduce production costs, and improve product market competitiveness.

[0122] Example 4: Wind farm unit operation optimization

[0123] Data collection and preprocessing, specifically:

[0124] Sensors are installed in the nacelle, hub, tower and other parts of each wind turbine generator set in the wind farm. The sensors include anemometer (measuring range 0m / s-50m / s, measuring accuracy ±0.5m / s), wind vane (measuring range 0°-360°, measuring accuracy ±5°), power sensor (measuring range 0kW-5000kW, measuring accuracy ±1%), speed sensor (measuring range 0rpm-30rpm, measuring accuracy ±0.1rpm), temperature sensor (measuring generator winding temperature, gearbox oil temperature, etc., range -40℃-150℃), vibration sensor (measuring vibration amplitude and frequency of key components of the unit). The data collection time interval is 10 minutes (for environmental parameters such as wind speed and wind direction) to 1 hour (for slowly changing parameters such as lubricating oil quality changes).

[0125] The collected data is transmitted to the data processing center, and the data cleaning module first checks the data. For abnormal wind speed data caused by icing or failure of the anemometer (such as the wind speed suddenly becomes zero or a maximum value), it is identified and corrected, and the average value of adjacent normal data or the predicted value based on historical data can be used to replace it; for abnormal readings of the power sensor, the rationality is judged in combination with the wind speed and rotation speed data. If it is unreasonable, it is corrected or deleted. Then the data is normalized, the wind speed data is normalized to the [0,1] interval, the power data is normalized to the [-1,1] interval, etc., to facilitate subsequent calculations.

[0126] Feature extraction and selection, specifically:

[0127] According to the wind power generation principle and the factors affecting the unit performance, the key features are determined. Wind speed, wind direction, unit power, generator speed, blade angle, gearbox oil temperature, ambient temperature, etc. are selected as features.

[0128] Due to the complex nonlinear relationship between wind speed and power, kernel principal component analysis (KPCA) is used for feature extraction. The principal components with a cumulative contribution rate of more than 92% are selected as the main features, reducing the original multiple feature dimensions to 3-4, reducing the data dimensions while retaining the main information.

[0129] Cluster analysis between micro-conditions and micro-intervals, specifically:

[0130] The density-based DBSCAN algorithm is used to cluster the operating conditions of wind turbines. According to the geographical location, climate conditions and operating characteristics of the wind farm, the neighborhood radius is set to an appropriate value (for example, 0.3-0.5 according to the distribution of wind speed and power data, which is the normalized value here), and the minimum number of sample points is 3-5 (determined according to the density of data points and the expected clustering effect). The DBSCAN algorithm is used to divide the operating conditions of wind turbines into different micro-conditions / micro-intervals, such as low wind speed start-up area, low wind speed power generation area, medium wind speed high-efficiency power generation area, high wind speed power-limited operation area, shutdown maintenance area, etc.

[0131] Use the silhouette coefficient to evaluate the clustering results. If the silhouette coefficient is not ideal, adjust the neighborhood radius and the minimum number of sample points and re-perform the clustering analysis until a reasonable clustering result is obtained. For example, if the data points in a cluster are found to be too scattered and the silhouette coefficient is low, the neighborhood radius can be appropriately reduced to make the cluster more compact; if some data points are incorrectly divided into different clusters, the minimum number of sample points can be adjusted to obtain a more accurate cluster boundary.

[0132] Micro-condition / micro-interval classification and modeling, specifically:

[0133] Use neural networks to build micro-condition / micro-interval classification models, use preprocessed and feature-selected data as input, and use micro-condition / micro-interval category labels as output to train the neural network model. By adjusting the structure of the neural network (such as the number of hidden layers and nodes) and training parameters (such as learning rate and number of iterations), the accuracy and generalization ability of the classification model can be improved.

[0134] For each micro-operating condition / micro-interval, a performance analysis model between the unit's power generation efficiency and operating parameters is established. For example, in the medium wind speed high-efficiency power generation area, a mathematical model between power generation efficiency and wind speed, blade angle, and generator speed is established, and multiple linear regression or nonlinear regression methods (if the data relationship presents nonlinear characteristics) can be used. A large amount of historical operating data is used to train and optimize the model so that the model can accurately predict the unit's power generation efficiency under different operating conditions.

[0135] Micro-condition / micro-interval analysis and application, specifically:

[0136] The distribution and correlation of operating parameters within each micro-condition / micro-interval are analyzed. In the low wind speed power generation area, it is found that the blade angle is positively correlated with the power generation. Properly increasing the blade angle can increase the power generation, but too large a blade angle may cause the vibration of the unit to increase. In the high wind speed power-limited operation area, the wind speed and power generation are no longer linearly related. At this time, the power output is mainly controlled by adjusting the blade angle and generator torque, while maintaining the stable operation of the unit.

[0137] Formulate optimization strategies based on the analysis results. In low wind speed power generation areas, when the wind speed increases, optimize the blade angle adjustment strategy based on the performance analysis model and real-time monitoring data to obtain the maximum power generation; in high wind speed power-limited operation areas, monitor the wind speed and unit vibration in real time, and reasonably adjust the blade angle and generator torque to ensure that the unit can maximize the power generation efficiency under the premise of safety and stability. At the same time, combined with real-time monitoring data, the performance analysis model is used to predict the power generation efficiency of the unit. When it is predicted that the power generation efficiency will decrease, check the equipment status of the unit in advance, such as whether the blades are damaged, whether the gearbox is poorly lubricated, etc., and take corresponding measures in time, such as repairing the blades, replacing lubricating oil, etc., to ensure the efficient and stable operation of the wind turbine and improve the overall power generation efficiency of the wind farm.

[0138] Example 5: Urban rail transit system operation optimization

[0139] Data collection and preprocessing, specifically:

[0140] Various sensors are installed in stations (including platforms, station halls, equipment rooms, etc.), trains (inside the carriages, front and rear of the trains, etc.) and along the tracks (track temperature, humidity sensors, etc.) of the urban rail transit system. The sensors include passenger flow sensors (counting passenger flow in and out of the station, real-time passenger flow in the carriages, with an accuracy of ±5%), train position sensors (determining the exact position of the train on the line, with an accuracy of ±10 meters), train speed sensors (measuring range 0km / h-160km / h, accuracy ±1km / h), power supply system parameter sensors (measuring voltage, current, power, etc., with voltage measurement accuracy of ±1%, current accuracy of ±2%, power accuracy of ±3%), ambient temperature sensors (measuring the temperature of stations and along the tracks, ranging from -20℃-40℃), and humidity sensors (measuring range 20%-90%). The data collection time interval is determined according to the changing characteristics of different parameters. Passenger flow data is collected every 1 minute during peak hours and every 5 minutes during non-peak hours; train speed and position data are collected in real time; and environmental parameters are collected every 10 minutes.

[0141] The collected data is transmitted to the data processing center, and the data cleaning submodule first cleans the data. For abnormal passenger flow data caused by passenger flow sensor failure (such as sudden maximum or minimum values), correction can be made, and the passenger flow data of the same period or adjacent stations can be referred to for smoothing; for abnormal readings of train speed sensors, rationality judgment is made in combination with the train operation schedule and adjacent position sensor data. If it is unreasonable, it is corrected or deleted. Then the data is normalized, for example, the passenger flow data is normalized to the [0,1] interval according to the maximum design passenger flow of the station, and the temperature data is normalized to the [-1,1] interval, etc., so that different types of data are comparable.

[0142] Feature extraction and selection, specifically:

[0143] According to the operation characteristics of urban rail transit system and the factors affecting service quality, key features are selected, including station passenger flow (divided into inbound, outbound and transfer passenger flow), train load rate (ratio of actual passenger flow in the carriage to rated passenger capacity), train running interval (the departure interval between two adjacent trains), train punctuality rate (the proportion of trains whose actual arrival time is within a certain range from the planned arrival time), power supply system energy consumption (power consumption per unit time), ambient temperature, humidity, etc.

[0144] Due to the complex nonlinear relationship between passenger flow and time, kernel principal component analysis (KPCA) is used for feature extraction. The principal components with a cumulative contribution rate of more than 85% are selected as the main features, reducing the original multiple feature dimensions to 4-6, which can reduce the amount of calculation and retain information that has an important impact on the operating status.

[0145] Cluster analysis between micro-conditions and micro-intervals, specifically:

[0146] The fuzzy clustering algorithm (FCM) is used to cluster the operating conditions of the urban rail transit system. According to the characteristics of the peak, flat and low peak periods of operation and different lines and stations, the fuzzy index m=2 (which can be adjusted according to the actual clustering effect) is set, and the number of cluster centers is 5 (for example, it can be divided into different operating conditions such as extreme peak, peak, flat, low peak, maintenance and overhaul, etc., which is determined according to the actual operating conditions). The operating conditions are divided into different micro conditions / micro intervals through the FCM algorithm.

[0147] Use the Calinski-Harabasz index to evaluate the clustering results. If the clustering effect is not good, adjust the fuzzy index m or the number of cluster centers and re-perform the clustering analysis until a satisfactory clustering result is obtained. For example, if it is found that the clustering distinction between peak and off-peak hours at some stations is not obvious, the clustering parameters can be appropriately adjusted to make different working conditions more clearly divided.

[0148] Micro-condition / micro-interval classification and modeling, specifically:

[0149] The decision tree algorithm is used to build a micro-condition / micro-interval classification model, with pre-processed and feature-selected data as input and micro-condition / micro-interval category labels as output to train the decision tree model. Pruning and other technologies are used to prevent model overfitting, improve the generalization ability of the model, and ensure that different operating conditions can be accurately identified.

[0150] For each micro-operating condition / micro-interval, a performance analysis model related to the quality of operation service is established. For example, in the peak operation micro-operating condition interval, a relationship model between train punctuality and train running interval, passenger flow, and power supply system stability is established, and methods such as multivariate linear regression or neural network can be used; in the low-peak operation micro-operating condition interval, a model is established between the power supply energy consumption per unit mileage and the operating status of the train running speed, carriage lighting, air conditioning and other equipment. A large amount of historical operation data is used to train and optimize the model so that the model can accurately reflect the relationship between the operating parameters and service quality indicators in each micro-operating condition / micro-interval.

[0151] Micro-condition / micro-interval analysis and application, specifically:

[0152] The distribution and correlation of operating parameters in each micro-operating condition / micro-interval are analyzed. During peak operation, it is found that the station passenger flow is positively correlated with the train load factor, and the train running interval has a significant impact on the load factor and punctuality rate; during off-peak operation, the train speed is positively correlated with the power supply energy consumption. Appropriately reducing the train speed can reduce energy consumption, but the minimum operating time requirement must be met.

[0153] Formulate optimization strategies based on the analysis results. During peak hours, dynamically adjust the train running interval according to the real-time passenger flow and train operation status, increase train formation or adopt large and small routes and other operation modes to improve transportation capacity and ensure train punctuality; during off-peak hours, reasonably optimize the train running speed, flexibly adjust the operation mode of station equipment (such as lighting, air conditioning, etc.) according to the passenger flow distribution, and reduce energy consumption. In addition, combined with real-time monitoring data, the performance analysis model is used to predict operational indicators such as train punctuality and energy consumption. When it is predicted that the punctuality rate may decrease or energy consumption will increase, take measures in advance, such as adjusting the signal system parameters, optimizing the equipment operation plan, etc., to ensure the efficient, safe, and energy-saving operation of the urban rail transit system and improve service quality and operational efficiency.

[0154] Embodiment six:

[0155] A classification and analysis system for micro-conditions / micro-intervals of complex systems, comprising:

[0156] The data acquisition and preprocessing module is configured to: acquire and preprocess the operating data of the complex system;

[0157] The micro-condition / micro-interval clustering analysis module is configured to: extract key features that can characterize the system operation status from the pre-processed data according to the operation principle and performance influencing factors of the complex system, and divide the corresponding data points into different clusters, each cluster representing a micro-condition / micro-interval;

[0158] The micro-condition / micro-interval classification module is configured to: determine a category label for each micro-condition / micro-interval according to the clustering result, determine the system operation characteristics and system performance indicators within the micro-condition / micro-interval by establishing a micro-condition / micro-interval classification model and a corresponding performance analysis model, and establish a correlation model between each micro-condition performance indicator and the micro-condition performance parameter;

[0159] The index optimization module is configured to: determine the distribution law, correlation and influence degree of the operating parameters on the system performance index according to the system operating characteristics within the micro-operating condition / micro-interval and save them;

[0160] The operation optimization module is configured to: determine the gap between the actual operating condition performance index of each micro-operating condition and the optimal index value, and determine the optimization technical route and method and the adjustment method of related parameters based on the correlation model of the optimal index value; through the index optimization of each micro-operating condition, the optimization of all operating indicators of the complex system is achieved.

[0161] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A classification and analysis method for micro-conditions / micro-intervals of complex systems, characterized in that: The following steps are involved: Acquire and pre-process the operating data of complex systems; According to the operating principles and performance influencing factors of complex systems, key features that can characterize the operating status of the system are extracted from the preprocessed data, and the data points corresponding to the key features are divided into different clusters, each cluster representing a micro-operating condition / micro-interval; According to the clustering results, a category label is determined for each micro-condition / micro-interval. By establishing a micro-condition / micro-interval classification model and a corresponding performance analysis model, the system operation characteristics and system performance indicators within the micro-condition / micro-interval are determined, and a correlation model between each micro-condition performance indicator and the micro-condition performance parameter is established. According to the system operating characteristics within the micro-operating conditions / micro-intervals, the distribution law and correlation of the operating parameters and the degree of influence on the system performance indicators are determined and saved; Determine the gap between the actual operating performance index of each micro-operating condition and the optimal index value, and determine the optimization technical route and method and the adjustment method of related parameters based on the correlation model of the optimal index value; through the optimization of the indicators of each micro-operating condition, the optimization of all operating indicators of the complex system can be achieved.

2. A classification and analysis method for complex system micro-conditions / micro-intervals as claimed in claim 1, characterized in that: Obtaining the operating data of the complex system, specifically: deploying a sensor network at a set location of the complex system to obtain operating data, the operating data at least includes the equipment operating parameters of the complex system, environmental parameters, energy parameters input into the complex system, and performance indicators of the output products.

3. The classification and analysis method of complex system micro-conditions / micro-intervals according to claim 1, characterized in that: Preprocessing specifically includes: processing the acquired operation data using at least one of data cleaning, interpolation, filtering and normalization.

4. The classification and analysis method of complex system micro-conditions / micro-intervals according to claim 1, characterized in that: From the preprocessed data, the features that can characterize the operating status of the system are extracted, and the features whose contribution rate to the operating status is greater than the set value are screened out as key features.

5. A classification and analysis method for complex system micro-conditions / micro-intervals as claimed in claim 4, characterized in that: Features that can characterize the operating status of the system are extracted, and based on the principal component analysis method, the correlation analysis method or the kernel principal component analysis method, features whose contribution rate to the operating status is greater than the set value are screened out as key features.

6. A classification and analysis method for complex system micro-conditions / micro-intervals as claimed in claim 1, characterized in that: Through clustering, the data points corresponding to the key features are divided into different cluster clusters. Specifically, a density-based clustering algorithm, a hierarchical clustering algorithm or a fuzzy clustering algorithm is used to perform cluster analysis on the operating conditions of the complex system according to the selected key features. By setting the neighborhood radius and the minimum number of sample points, the data during the operation of the complex system is divided into different cluster clusters. Each cluster represents a micro-condition / micro-interval, reflecting different operating condition intervals of the system operation process.

7. A classification and analysis method for complex system micro-conditions / micro-intervals as claimed in claim 6, characterized in that: Evaluate the clustering results and adjust the parameters in the clustering algorithm based on the evaluation results.

8. The classification and analysis method of complex system micro-conditions / micro-intervals according to claim 1, characterized in that: Complex systems include energy and power systems, chemical production systems, steel manufacturing systems, wind power generation systems or urban rail transit systems.

9. A classification and analysis method for complex system micro-conditions / micro-intervals as claimed in claim 1, characterized in that: Using multiple linear regression or neural network methods, a mathematical model between system performance indicators and operating parameters is established, the system operating characteristics and system performance indicators within micro-conditions / micro-intervals are determined, and an association model between each micro-condition performance indicator and micro-condition performance parameter is established.

10. A system for implementing the complex system micro-condition / micro-interval classification method according to any one of claims 1 to 9, characterized in that: include: The data acquisition and preprocessing module is configured to: acquire and preprocess the operating data of the complex system; The micro-condition / micro-interval clustering analysis module is configured to: extract key features that can characterize the system operation status from the pre-processed data according to the operation principle and performance influencing factors of the complex system, and divide the corresponding data points into different clusters, each cluster representing a micro-condition / micro-interval; The micro-condition / micro-interval classification module is configured to: determine a category label for each micro-condition / micro-interval according to the clustering result, determine the system operation characteristics and system performance indicators within the micro-condition / micro-interval by establishing a micro-condition / micro-interval classification model and a corresponding performance analysis model, and establish a correlation model between each micro-condition performance indicator and the micro-condition performance parameter; The index optimization module is configured to: determine the distribution law, correlation and influence degree of the operating parameters on the system performance index according to the system operating characteristics within the micro-operating condition / micro-interval and save them; The operation optimization module is configured to: determine the gap between the actual operating condition performance index of each micro-operating condition and the optimal index value, and determine the optimization technical route and method and the adjustment method of related parameters based on the correlation model of the optimal index value; through the index optimization of each micro-operating condition, the optimization of all operating indicators of the complex system is achieved.

Images