Full life cycle management system and method of gas-steam combined cycle unit

Abstract

The invention relates to the technical field of equipment monitoring and control, and discloses a full life cycle management system and method for a gas-steam combined cycle unit, and the method comprises the steps: firstly constructing a typical working condition parameter database and a unit performance prediction model, and constructing a first performance fingerprint database; meanwhile, obtaining historical operation data to construct a second performance fingerprint database; generating a first performance evaluation report and a second performance evaluation report based on the first performance fingerprint database and the second performance fingerprint database; identifying abnormal working conditions according to the two performance evaluation reports, and generating an abnormal working condition fingerprint database and an abnormal working condition processing strategy database; and finally, performing abnormality diagnosis on the unit, and generating a working condition abnormality handling scheme of the unit. According to the invention, comprehensive performance evaluation and abnormal working condition diagnosis can be carried out on the gas-steam combined cycle unit, intelligent management of the whole life cycle of the unit is realized, and the economical efficiency and reliability of unit operation are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of equipment monitoring and control, and more specifically, to a full life cycle management system and method for a gas-steam combined cycle unit. Background Art

[0002] During the operation of gas-steam combined cycle units, the performance evaluation and abnormal condition diagnosis of the units are the key to ensuring the safe, economical and efficient operation of the units. At present, there are some patents involving intelligent dispatching and full life cycle management of equipment in thermal power plants and wind farms.

[0003] The Chinese patent with the authorization announcement number CN102541036B discloses an intelligent dispatching system for coal-fired power plants. The system realizes the full life cycle management and intelligent dispatching of coal-fired power plants through modules such as online coal panning, intelligent stacking, intelligent coal blending and coal loading calculation, ensuring the safety, economy and environmental protection of boiler units under mixed combustion conditions. However, this patent is mainly aimed at coal-fired boilers in thermal power plants and does not involve gas-steam combined cycle units. However, there are significant differences between gas-steam combined cycle units and coal-fired boilers in terms of operating characteristics, performance indicators, etc., and it is necessary to build a special performance evaluation and abnormal diagnosis model.

[0004] The Chinese patent with publication number CN106228262A discloses a system and method for full life cycle management of wind turbines. The system adopts data flow management, organically combines ERP, CRM, PDM and other systems through data flow, and establishes a full life cycle management system for wind turbine operation, which can provide wind farm owners with a complete information management platform and economical operation and maintenance plan reference. However, this patent is mainly aimed at wind turbines, which are quite different from gas-steam combined cycle units in terms of unit structure, operating conditions, performance indicators, etc., and cannot be directly applied to the performance evaluation and abnormal condition diagnosis of gas-steam combined cycle units.

[0005] In summary, the prior art lacks an effective method specifically for the full life cycle management of gas-steam combined cycle units. Summary of the invention

[0006] In order to overcome the above-mentioned defects of the prior art, the present invention provides a full life cycle management system and method for a gas-steam combined cycle unit, constructs a performance fingerprint library suitable for the operating conditions of the gas-steam combined cycle unit, and accurately evaluates the performance level of the unit through typical operating condition analysis and historical data mining; at the same time, an abnormal operating condition diagnosis model is established to realize online early warning and diagnostic decision-making of abnormal operating conditions of the unit, so as to ensure safe, economical and efficient operation of the unit.

[0007] To achieve the above object, the present invention provides the following technical solutions: The full life cycle management method of gas-steam combined cycle units includes: Constructing a typical operating parameter database and a unit performance prediction model, and constructing a first performance fingerprint library based on the typical operating parameter database and the unit performance prediction model; acquiring second historical operating data, and constructing a second performance fingerprint library based on the second historical operating data; Generate a first performance evaluation report based on the first performance fingerprint library and the preset performance evaluation benchmark; generate a second performance evaluation report based on the first performance fingerprint library and the second performance fingerprint library; identify abnormal working conditions according to the first performance evaluation report and the second performance evaluation report, and generate an abnormal working condition fingerprint library and an abnormal working condition handling strategy library; Provide online warning of abnormal operating conditions for the unit and output abnormal diagnosis results; match the abnormal operating condition handling strategy library based on the abnormal diagnosis results to generate the unit's abnormal operating condition handling plan.

[0008] Furthermore, the construction of a typical operating condition parameter database includes: Get gas-steam combined cycle unitn 1 The first level typical working condition parameters under the typical working condition; n 1 is the total number of typical working conditions; Based on The first-level typical operating condition parameters under the typical operating condition are calculated to obtain the The second level typical working condition parameters under the typical working condition; 1≤ ≤n 1 ; The first The first level typical operating parameters and the second level typical operating parameters under the typical operating conditions The second level typical operating condition parameters under the typical operating condition constitute the Record the working parameters of typical working conditions; The operating parameters of all typical operating conditions are recorded to form a typical operating condition parameter database; the typical operating condition parameter database includes n 1 Group operating parameter records.

[0009] Furthermore, the constructing of the unit performance prediction model includes: Get History T 1 The historical operation data of the gas-steam combined cycle unit within the time period is marked as the first historical operation data; and a unit performance prediction model is constructed and trained based on the first historical operation data; The constructing of the first performance fingerprint library comprises: According to typical operating parameters and unit performance prediction model, we can get n 1 Predictive performance indicators under typical working conditions; Based on the typical operating condition parameter database The working parameter records of the typical working conditions and the The predicted performance index under typical working conditions is generated Performance fingerprint under typical working conditions; The performance fingerprints under all typical working conditions constitute the first performance fingerprint library.

[0010] Furthermore, the second historical operation data is historical T 2 Historical operating data of the gas-steam combined cycle unit within the time period, T 2 ≠T 1 The second historical operation data includes n 4 Group historical operating parameter records, n 4 is the total number of historical working conditions; The second historical operation data includes a second historical operating condition parameter and a second historical performance index; The constructing of the second performance fingerprint library comprises: generating a historical performance fingerprint based on a second historical operating condition parameter and a second historical performance index; and forming all the historical performance fingerprints into a second performance fingerprint library.

[0011] Further, generating the first performance evaluation report includes: Traverse the performance fingerprint of each typical working condition in the first performance fingerprint library and extract the The predicted performance index of the typical working condition is calculated. The predicted performance index of a typical working condition is consistent with the preset Deviation from the performance evaluation benchmark for a typical operating condition; For The deviation of all the prediction performance indicators of the typical working condition is weighted averaged to obtain the The comprehensive performance deviation score of typical working conditions; The comprehensive performance deviation scores of all typical working conditions are compared with a preset deviation threshold, typical working conditions that are higher than the deviation threshold are marked as weak working conditions, and a first performance evaluation report is generated.

[0012] Further, generating the second performance evaluation report includes: Traverse all historical performance fingerprints in the second performance fingerprint library, perform similarity matching with the first performance fingerprint library, and identify the operating conditions of the unit; the operating conditions of the unit are divided into typical operating conditions and atypical operating conditions; Based on the identification results of the unit's operating conditions, analyze the unit's cumulative operating time and start-stop frequency under various typical and atypical operating conditions; Based on the cumulative running time and start-stop frequency, the effective utilization rate of the computer group under various typical and atypical working conditions; analyze T 3The changing trend of the effective utilization rate of the unit under various atypical operating conditions within a time period.

[0013] Further, the identifying the working condition of the unit includes: Calculate the second performance fingerprint library The historical performance fingerprint and the first performance fingerprint library Similarity of performance fingerprints , get the similarity sequence; 1≤ ≤n 2 , 1≤ ≤n 3 ;n 2 is the number of historical performance fingerprints in the second performance fingerprint library, n 3 is the number of performance fingerprints in the first performance fingerprint library, n 3 =n 1 , n 2 =n 4 ; If the similarity sequence is the maximum value, and ≥ , then it is considered that The historical operating condition corresponding to the historical performance fingerprint belongs to Typical working conditions corresponding to each performance fingerprint; is the preset similarity threshold; If the similarity sequence is the maximum value, and < , then the The historical operating conditions corresponding to the historical performance fingerprints are marked as atypical operating conditions.

[0014] Furthermore, the identifying of abnormal operating conditions and generating an abnormal operating condition fingerprint library and an abnormal operating condition handling strategy library include: According to the comprehensive performance deviation score of typical working conditions, the typical working conditions are divided into four abnormal levels: normal, slightly abnormal, moderately abnormal, and severely abnormal; the typical working conditions with slightly abnormal, moderately abnormal, and severely abnormal are marked as abnormal working conditions; the normal typical working conditions are marked as stable working conditions; Extracting performance indicator characteristics of abnormal working conditions; generating an abnormal working condition fingerprint library according to working condition parameters, performance indicator characteristics and abnormal level of abnormal working conditions; For abnormal operating conditions of different abnormal levels, abnormal operating condition handling strategies are configured according to the first performance evaluation report and the second performance evaluation report, and an abnormal operating condition handling strategy library is generated.

[0015] Furthermore, the online early warning of abnormal operating conditions of the unit includes: Collect the real-time operating parameters of the unit, and obtain the real-time predicted performance indicators based on the real-time operating parameters and the unit performance prediction model; Generate the real-time performance fingerprint of the unit based on the real-time operating parameters and real-time predicted performance indicators; Calculate the similarity between the real-time performance fingerprint and each abnormal working condition fingerprint in the abnormal working condition fingerprint library to generate a second similarity sequence SI; Find the maximum value from the second similarity sequence SI, recorded as SI A , A represents the abnormal condition fingerprint with the highest similarity to the real-time performance fingerprint in the abnormal condition fingerprint library; SI A and the preset abnormality judgment threshold θ 1 For comparison, if SI A <θ 1 , the current real-time operating condition is determined to be a stable operating condition; otherwise, it is determined to be an abnormal operating condition, and the abnormal level of the current real-time operating condition is determined according to the abnormal level of the abnormal operating condition corresponding to the abnormal operating condition fingerprint A.

[0016] A full life cycle management system for a gas-steam combined cycle unit, which is used to implement the full life cycle management method for the gas-steam combined cycle unit, and the system includes: Fingerprint library construction module: used to construct a typical operating condition parameter database and a unit performance prediction model, and to construct a first performance fingerprint library based on the typical operating condition parameter database and the unit performance prediction model; to obtain second historical operation data, and to construct a second performance fingerprint library based on the second historical operation data; Performance evaluation module: generates a first performance evaluation report based on the first performance fingerprint library and a preset performance evaluation benchmark; generates a second performance evaluation report based on the first performance fingerprint library and the second performance fingerprint library; Working condition diagnosis module: identifies abnormal working conditions according to the first performance evaluation report and the second performance evaluation report, and generates an abnormal working condition fingerprint library and an abnormal working condition disposal strategy library; Online analysis module: used to provide online warning of abnormal operating conditions of the unit and output abnormal diagnosis results; according to the abnormal diagnosis results, match the abnormal operating condition handling strategy library to generate the unit's abnormal operating condition handling plan.

[0017] An electronic device comprises a memory, a central processing unit and a computer program stored in the memory and executable on the central processing unit, wherein the central processing unit implements the above-mentioned full life cycle management method of the gas-steam combined cycle unit when executing the computer program.

[0018] A computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed, implements the above-mentioned full life cycle management method for a gas-steam combined cycle unit.

[0019] Compared with the prior art, the present invention has the following beneficial effects: The gas-steam combined cycle unit life cycle management system and method provided by the present invention can build a unit performance fingerprint library based on the unit's typical operating parameters and historical operating data, and achieve a comprehensive characterization of the unit's operating conditions and performance. By generating a performance evaluation report, the present invention can accurately evaluate the performance of the unit under different operating conditions, identify the shortcomings and bottlenecks of the unit's operation, and provide a decision-making basis for the unit's optimized operation and maintenance strategy. At the same time, the present invention can also perform online early warning and diagnosis of abnormal operating conditions of the unit, quickly identify the abnormal operating condition level of the unit through matching of the abnormal operating condition fingerprint library, and automatically generate the corresponding abnormal operating condition disposal plan, providing intelligent auxiliary decision support for the safe and efficient operation of the unit. In addition, by analyzing the cumulative operating time and start-stop frequency of the unit under different typical and atypical operating conditions, the present invention can evaluate the adaptability of the unit to different operating conditions and provide guidance for the scheduling optimization of the unit's operating conditions. In summary, the present invention makes full use of the massive operating data of the unit throughout its life cycle, combines data-driven modeling analysis and diagnostic decision-making technology, and realizes the intelligent perception, evaluation, early warning, diagnosis and optimization of the gas-steam combined cycle unit, which can significantly improve the management level and operating efficiency of the unit, reduce the risk of unplanned downtime, and lay the foundation for the economical operation of the unit throughout its life cycle. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0021] Figure 1 It is a principle flow chart of the full life cycle management method of the gas-steam combined cycle unit in the present invention; Figure 2 A flow chart of a method for constructing a typical operating condition parameter database in the full life cycle management method of a gas-steam combined cycle unit of the present invention; Figure 3 A flow chart of a method for constructing a first performance fingerprint library in the full life cycle management method of a gas-steam combined cycle unit of the present invention; Figure 4 A flow chart of a method for constructing a second performance fingerprint library in the full life cycle management method of a gas-steam combined cycle unit of the present invention; Figure 5A flow chart of a method for generating a first performance evaluation report in a full life cycle management method for a gas-steam combined cycle unit of the present invention; Figure 6 A flow chart of a method for generating a second performance evaluation report in the full life cycle management method of a gas-steam combined cycle unit of the present invention; Figure 7 It is a flow chart of a method for online early warning of abnormal operating conditions of a unit in the full life cycle management method of a gas-steam combined cycle unit of the present invention; Figure 8 It is a functional module diagram of the full life cycle management system of the gas-steam combined cycle unit in the present invention. DETAILED DESCRIPTION

[0022] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only 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 ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0023] Example 1 See also Figure 1 As shown, this embodiment provides a full life cycle management method for a gas-steam combined cycle unit, including: Step S1000, obtaining typical operating parameters of the gas-steam combined cycle unit and constructing a typical operating parameter database; constructing a unit performance prediction model, and constructing a first performance fingerprint library based on the typical operating parameter database and the unit performance prediction model; obtaining second historical operation data, and constructing a second performance fingerprint library based on the second historical operation data; Furthermore, step S1000 includes: Step S1100, obtaining typical operating parameters of the gas-steam combined cycle unit and constructing a typical operating parameter database, wherein the typical operating parameters include first-level typical operating parameters and second-level typical operating parameters; Furthermore, if Figure 2 As shown, step S1100 includes: Step S1110, obtaining gas-steam combined cycle unit n 1 The first level typical working condition parameters under the typical working condition; n 1 is the total number of typical working conditions; Step S1120, based on The first-level typical operating condition parameters under the typical operating condition are calculated to obtain the The second level typical working condition parameters under the typical working condition; 1≤ ≤n1 ; Step S1130: The first level typical operating parameters and the second level typical operating parameters under the typical operating conditions The second level typical operating condition parameters under the typical operating condition constitute the Record the working parameters of typical working conditions; Step S1140: Record the operating parameters of all typical operating conditions to form a typical operating condition parameter database; the typical operating condition parameter database includes n 1 Group operating parameter records.

[0024] Specifically, the purpose of step S1100 is to construct a database that comprehensively describes the performance of the gas-steam combined cycle unit under various typical operating conditions. Each record in the database contains the key parameters of the unit under a specific operating condition, and these parameters are divided into two levels: The first-level typical operating parameters refer to those parameters that are easy to measure directly, such as unit load (reflecting the unit's power output level), ambient temperature (affecting the gas turbine intake conditions) and fuel characteristics (such as the calorific value and composition of natural gas). Although these parameters are external input conditions, they have a significant impact on the performance of the unit. Unit load refers to the set value or actual value of the unit's power generation, and is one of the key parameters for measuring the unit's operating status. Under different loads, the unit's combustion, flow, heat transfer and other operating conditions vary greatly. Ambient temperature refers to the atmospheric temperature at the unit's geographical location, measured in °C. Ambient temperature significantly affects the gas turbine's intake density and pressure ratio, and thus affects the operating conditions of the combustion chamber and steam turbine. Fuel characteristics refer to the type and composition of the fuel used by the unit. The calorific value, sulfur content, ash content and other characteristics of different fuels vary greatly, which directly affects the operating conditions of the combustion chamber and exhaust emissions. These first-level operating parameters usually come from design manuals, experimental records or DCS (distributed control system) monitoring data, and are easy to obtain directly. However, they only reflect the "surface" state of the unit and it is difficult to fully characterize the internal thermal system conditions.

[0025] The second-level typical operating parameters refer to those parameters that are difficult to measure directly but are more critical to judging the internal operating status of the unit, such as the combustion chamber outlet temperature, turbine inlet temperature and condenser vacuum. These parameters need to be estimated based on the first-level parameters through algorithms such as thermodynamic calculations. The combustion chamber outlet temperature refers to the average temperature of the flue gas at the outlet of the gas turbine combustion chamber, in K or °C. This temperature determines the inlet temperature of the gas turbine and is a key parameter affecting the cycle efficiency and turbine life. The turbine inlet temperature refers to the temperature of the main steam at the inlet of the high-pressure cylinder of the turbine, in °C. The inlet steam temperature directly affects the cycle efficiency and condensation volume of the turbine. The condenser vacuum refers to the absolute pressure inside the condenser, in kPa. The vacuum determines the back pressure of the turbine exhaust, affecting the cycle output and energy utilization level.

[0026] The second-level operating parameters are difficult to measure directly and need to be derived from the first-level parameters through algorithms based on thermodynamics and heat transfer knowledge. Taking the combustion chamber outlet temperature as an example, it can be solved based on the combustion chamber energy balance equation: In the formula, The heat input to the fuel, is the flue gas mass flow rate, is the specific heat of flue gas at constant pressure, and are the compressor outlet temperature and the combustion chamber outlet temperature, respectively. is the heat loss of the combustion chamber. From the compressor characteristic curve, we can get and , combined with the known fuel lower calorific value, excess air coefficient, etc., the combustion chamber outlet temperature can be solved Other second-level parameters can also be derived in a similar way.

[0027] Step S1130 combines the primary parameters and secondary parameters under each typical operating condition to form a complete operating condition parameter record. Each record comprehensively describes the thermal system state of the unit under the operating condition "from the outside to the inside". Step S1140 1 The parameter records of typical operating conditions are summarized to build a typical operating condition parameter database. This database covers a comprehensive description of the unit's operating conditions in different operating scenarios, but it does not require that each operating condition actually occurs. It is more like an "encyclopedia" of unit performance, used to support subsequent performance prediction, evaluation and optimization tasks.

[0028] The typical operating condition parameter database covers the complete performance spectrum of the unit from rated operating conditions to special operating conditions, without missing any possible operating scenarios. This provides a solid operating condition foundation for full life cycle management. The performance of different typical operating conditions reflects the adaptability of the unit under different operating conditions. They constitute a set of multi-dimensional and three-dimensional performance evaluation scales. By obtaining the first-level easy-to-measure parameters and then deriving the second-level deep parameters, the internal and external operating conditions of the unit can be fully described. Summarizing the records of various typical operating conditions into a database can provide ideal reference and sample support for subsequent tasks such as unit performance prediction, evaluation, and optimization, making performance management more scientific, accurate, and comprehensive.

[0029] Step S1200, constructing a unit performance prediction model; Further, step S1200 includes: Step S1210, obtain the history T 1 The historical operation data of the gas-steam combined cycle unit within the time period is marked as the first historical operation data; the first historical operation data includes the first historical operating condition parameter and the first historical performance index; Step S1220: construct and train a unit performance prediction model based on the first historical operation data.

[0030] Specifically, the unit performance prediction model is a mathematical model based on a machine learning algorithm, which is used to predict the performance indicators of a gas-steam combined cycle unit under given operating parameters. The model learns a large amount of historical operating data and explores the intrinsic relationship between operating parameters and performance indicators to achieve quantitative performance prediction. The purpose of building this model is to accurately estimate the performance of the unit in the absence of a physical mechanism model, and to provide a quantitative basis for performance evaluation, optimization and management. The input of the model is the operating parameters representing the operating status of the unit, such as load, ambient temperature, fuel characteristics, etc.; the output is the key indicators for measuring the performance of the unit, such as power generation efficiency, fuel consumption rate, etc. By training and optimizing the model parameters, the model can have strong nonlinear fitting and generalization prediction capabilities.

[0031] In step S1210, a certain historical time range (denoted as T) is obtained from data sources such as the unit's operation records, monitoring system, and performance test reports. 1 ) as samples for training the performance prediction model. The historical operating data consists of two parts: operating parameters and performance indicators. The first historical operating parameters describe the operating conditions of the unit at that time, such as load, ambient temperature, fuel calorific value, etc., which correspond to the input variables of the model; the first historical performance indicators reflect the actual performance of the unit under the corresponding conditions, such as power generation efficiency, coal consumption rate, etc., which correspond to the output variables of the model, also called labels or target values. The operating parameters and performance indicators at the same time point are combined into data pairs to form a complete training sample.

[0032] Step S1220 uses the historical operating data obtained to build and train the unit performance prediction model. The historical data is randomly divided into a training set and a test set. The training set is used to train and optimize the model parameters, and the test set is used to evaluate the performance and generalization ability of the model. A variety of machine learning algorithms can be used, such as support vector machines, random forests, neural networks, etc., and the appropriate algorithm is selected according to the characteristics of the data and the prediction requirements. Taking the neural network as an example, the network structure parameters such as the number of input layer nodes (equal to the number of operating parameters), the number of hidden layers and the number of nodes per layer, and the number of output layer nodes (equal to the number of performance indicators) are determined, and hyperparameters such as the number of training iterations, learning rate, and regularization parameters are set. Input the training set sample, iteratively optimize the model parameters, so that the predicted value is as close to the actual value as possible. Then use the test set sample to test the model performance, and calculate the mean square error, determination coefficient and other evaluation indicators. Optimize hyperparameters through methods such as cross-validation to improve the robustness and generalization ability of the model.

[0033] After training and optimization, the performance prediction model can establish a nonlinear mapping relationship from operating parameters to performance indicators. When new operating parameters are input, the model can more accurately predict the unit performance indicators under the operating conditions without the need for tedious physical modeling and calculations. Since the training data covers a long period of actual operating conditions, the model can characterize the performance characteristics of the unit under different operating conditions, and the prediction results are more convincing and credible.

[0034] Step S1300, constructing a first performance fingerprint library; Furthermore, if Figure 3 As shown, step S1300 includes: Step S1310, according to typical operating parameters and unit performance prediction model, obtain n 1 Predictive performance indicators under typical working conditions; Step S1320, based on the typical operating condition parameter database The working parameter records of the typical working conditions and the The predicted performance index under typical working conditions is generated Performance fingerprint under typical working conditions; Step S1330: The performance fingerprints under all typical working conditions are combined into a first performance fingerprint library.

[0035] Specifically, the construction of the performance fingerprint library is to systematically and comprehensively describe the performance characteristics of the unit under different operating conditions. The first performance fingerprint library is generated based on the typical operating condition parameter database and the unit performance prediction model. It reflects the multi-operating condition performance of the unit under the design and ideal state. It is an inherent attribute of the unit, so it is called the "first performance fingerprint". This fingerprint library provides an important reference benchmark for quantitatively evaluating the actual operating performance of the unit.

[0036] Load the typical operating condition parameter database constructed in step S1100 and traverse n 1 Typical working condition. The corresponding operating condition parameters are input into the unit performance prediction model trained in step S1220 to obtain the predicted performance index under the operating condition. The predicted performance index usually includes power generation efficiency, fuel consumption rate, pollutant emission level, etc., which reflects the comprehensive performance of the unit under the operating condition and is the nonlinear mapping result from the operating condition parameters to the performance index. Repeat this process until all n 1 Predictive performance indicators under typical working conditions.

[0037] It should be noted that before inputting typical operating parameters into the model, necessary data preprocessing must be performed, including data cleaning, normalization, and outlier processing. Data cleaning is to remove noise, outliers, and obviously erroneous records in the original data to improve data quality. Outlier processing is to identify and correct abnormal data based on certain rules (such as the 3σ criterion) to return it to normal levels. Data normalization is to uniformly map operating parameters of different dimensions to the [0,1] interval to eliminate the dimension effect and facilitate model learning. The above preprocessing can improve the accuracy and robustness of model prediction.

[0038] The performance fingerprint is a combination of operating parameters and performance indicators formed according to a certain logic, just as a fingerprint pattern is composed of fine lines. The operating parameter records of a typical operating condition and the corresponding predicted performance indicators are matched and integrated, and the key information is extracted, and they are expressed and stored in a certain data structure such as vectors and matrices to generate the performance fingerprint under the operating condition. On the one hand, the performance fingerprint inherits the basic information of the operating condition parameters and reflects the operating condition of the unit; on the other hand, the performance fingerprint incorporates the information of the performance indicators and reflects the performance level of the unit under the operating condition. Therefore, the performance fingerprint can be regarded as a performance portrait of the unit "from the outside to the inside" under a certain operating condition.

[0039] All typical operating conditions (a total of n) generated in step S1320 are 1 The performance fingerprints of different types are summarized to form a complete performance fingerprint library. The first performance fingerprint library covers all the ideal performance characteristics of the unit from rated conditions to special conditions, providing important basic data support for full life cycle performance management. The performance fingerprints of different typical conditions reflect the ability of the unit to adapt to different conditions, reveal the inherent laws of its performance, and help manufacturers optimize the design and improve the unit. The collection of all performance fingerprints constitutes a comprehensive and systematic performance evaluation framework, which is convenient for quantitative analysis and comparison of the performance of the unit in different states, laying the foundation for subsequent performance evaluation, fault diagnosis, predictive maintenance and other tasks.

[0040] By constructing the first performance fingerprint library, the following beneficial effects can be obtained: 1. Provides a set of ideal performance reference benchmarks. The first performance fingerprint library is used as the "gold standard" to measure the actual operating performance of the unit. Through comparative analysis, the performance level of the unit can be quantitatively evaluated and performance anomalies and deterioration trends can be discovered in a timely manner.

[0041] 2. Reveals the inherent laws of unit performance. The correlation analysis between different fingerprints helps to explore the inherent relationship between operating parameters and performance indicators, and deeply understand the key factors affecting unit performance.

[0042] 3. Supports performance evaluation and optimization. The performance fingerprint library can be used to conduct multi-dimensional and multi-angle performance comparisons, identify deficiencies in design and operation, provide a practical basis for unit optimization and improvement, and improve unit efficiency and reliability.

[0043] 4. It is easy to build a performance prediction model. Using the performance fingerprint library as a model training sample that combines mechanism and data drive can significantly improve the model's fitting and extrapolation capabilities, making performance prediction more accurate and comprehensive.

[0044] In short, the first performance fingerprint library records in detail the performance characteristics of the unit under the design and ideal state, which is a concentrated reflection of the unit's innate endowment. Establishing and making full use of the performance fingerprint library is of great significance for in-depth understanding of the unit's performance laws, optimizing design and transformation plans, and quantitatively evaluating operating performance. It is an important cornerstone for achieving refined management and intelligent operation and maintenance of the unit.

[0045] Step S1400, constructing a second performance fingerprint library; Furthermore, if Figure 4 As shown, step S1400 includes: Step S1410, obtaining history T 2 The historical operation data of the gas-steam combined cycle unit within the time period is marked as the second historical operation data; the second historical operation data includes the second historical operating condition parameter and the second historical performance index; the second historical operation data includes n 4 Group historical operating parameter records, n 4 is the total number of historical operating conditions; T 2 ≠T 1 ; Step S1420, generating a historical performance fingerprint based on the second historical operating condition parameter and the second historical performance index; Step S1430: All historical performance fingerprints are combined into a second performance fingerprint library.

[0046] Specifically, the second performance fingerprint library is constructed based on the actual operation data of the unit, reflecting the actual performance of the unit in different historical periods, and is dynamically evolving. Compared with the first performance fingerprint library, the second performance fingerprint library reflects more of the impact of factors such as the external environment, equipment status, and personnel operation on the performance of the unit, so it is called the "second performance fingerprint". The purpose of building this fingerprint library is to explore the historical evolution of unit performance through big data analysis, so as to facilitate tasks such as performance evaluation, trend prediction, and abnormal diagnosis.

[0047] The collection time span of historical operation data is T 2 The collection time T of the first historical operation data used to train the unit performance prediction model 1 Different, so that the data source has diversity in the time dimension. The operating data within this time period is obtained from the unit's DCS system, performance monitoring system and other platforms as the original material for building the historical performance fingerprint library. Similar to the first performance fingerprint library, the second historical operation data also consists of two parts: operating parameters and performance indicators. The difference is that the second historical operating parameters reflect the operating condition changes experienced by the unit during actual operation, including both design conditions and non-design conditions, such as operating data under different loads and different environmental conditions. The second historical performance indicators reflect the actual performance level achieved by the unit under these actual conditions, which may deviate from its theoretical performance. Pair the operating parameters and performance indicators at the same time to form a complete set of historical operating parameter records. If the sampling frequency of historical data is 1 hour, 24 sets of records will be generated in one day.

[0048] The types of operating parameters and performance indicators in the second historical operating data are similar to those in the first performance fingerprint library, and are also divided into first-level parameters that are directly measured and second-level parameters that are indirectly calculated. However, due to factors such as missing data and noise interference, some first-level parameters may not be directly obtained and need to be estimated using reasonable interpolation and fitting methods. In addition, there may be outliers or inconsistencies in historical data, and a systematic data cleaning process must be adopted, including data screening, noise elimination, data repair, etc., to ensure the quality and reliability of the data.

[0049] For each set of records in the second historical operation data, its operating parameters and performance indicators are feature extracted and fused to generate the corresponding historical performance fingerprint. The content and form of the fingerprint are similar to those of the first performance fingerprint library, and can usually be expressed in data structures such as vectors, matrices, and time series. However, in the process of generating fingerprints, the following issues need to be focused on: (1) Working condition clustering: Since there are many types of historical working conditions, clustering the performance fingerprints of similar working conditions into one category can avoid the fingerprint library being too complicated. Clustering algorithms such as K-means and DBSCAN can be used to automatically divide historical working conditions into several typical categories.

[0050] (2) Missing value processing: Different from the ideal data integrity in the first performance fingerprint library, historical data may have missing parameters. For fingerprints with fewer missing values, reasonable interpolation methods can be used to complete them; for fingerprints with serious missing values, they should be removed or marked to avoid introducing errors.

[0051] (3) Statistical feature extraction: In the first performance fingerprint library, only one performance fingerprint is generated for each typical working condition. However, in historical data, the same working condition may appear multiple times. Therefore, it is necessary to extract the statistical features of these performance fingerprints, such as mean, variance, peak value, etc., as representative indicators of performance under this working condition.

[0052] The historical performance fingerprint reflects the performance level of the unit under different operating conditions in actual operation and contains a lot of valuable information. The historical performance fingerprints generated at all time points are summarized and organized according to time and operating conditions to form a complete second performance fingerprint library. This fingerprint library has the following characteristics: (1) Dynamicity: Different from the static characteristics of the first performance fingerprint library, the second performance fingerprint library is continuously updated over time and can reflect the changing trend of unit performance in real time.

[0053] (2) Diversity: The historical performance fingerprint covers various actual operating conditions since the unit was put into operation, reflecting the adaptability and performance of the unit under different conditions, and providing rich samples for comprehensive evaluation of unit performance.

[0054] (3) Correlation: There is often a certain correlation between the performance fingerprints of different periods and different operating conditions, which contains the inherent laws of unit performance evolution. Through correlation analysis and data mining, the influencing factors and deterioration mechanisms of performance can be revealed.

[0055] The benefits of building a second performance fingerprint library are: 1. Automatically evaluate performance. Using the historical performance fingerprint library, a performance evaluation model can be developed to automatically determine the current performance level by calculating the similarity between the actual working conditions and the historical fingerprints, without the need for manual table lookup.

[0056] 2. Reveal the law of performance evolution. Trend analysis of performance fingerprints in different periods helps to understand the general law of unit performance changes, early warning of performance deterioration, and optimization of operation and maintenance strategies.

[0057] 3. Diagnose abnormal performance conditions. Cluster historical performance fingerprints into different conditions and extract performance benchmarks for each condition. When the measured performance deviates significantly from the benchmark, it can be determined to be abnormal, providing clues for fault diagnosis.

[0058] 4. Accumulate optimization control knowledge. Under different operating conditions, sensitivity analysis of performance indicators can be performed to obtain key control variables that affect unit performance, which can be used to guide personnel to optimize control or develop automatic optimization systems.

[0059] 5. Enrich the performance portrait of the unit throughout its life cycle. As time goes by, the historical performance fingerprint library will become richer and richer, covering all stages of the unit from commissioning to retirement. Examining these performance fingerprints from a macro perspective can provide insight into the full life cycle characteristics of the unit's performance and provide data support for asset management and transformation decisions.

[0060] The above steps build a complete second performance fingerprint library by analyzing the historical operation big data of the unit. Compared with the first performance fingerprint library under ideal conditions, this fingerprint library is closer to the actual operation status of the unit and dynamically reflects the performance of the unit under different periods and conditions. Through horizontal comparison and vertical analysis of fingerprints, the unit performance can be quantitatively evaluated, the law of performance evolution can be revealed, and the foundation for subsequent performance management and optimization can be laid.

[0061] Step S2000, based on the first performance fingerprint library and the preset performance evaluation benchmark, generate a first performance evaluation report; based on the first performance fingerprint library and the second performance fingerprint library, generate a second performance evaluation report; according to the first performance evaluation report and the second performance evaluation report, identify abnormal working conditions from typical working conditions, and generate an abnormal working condition fingerprint library and an abnormal working condition handling strategy library; Furthermore, step S2000 includes: Step S2100, generating a first performance evaluation report based on the first performance fingerprint library and a preset performance evaluation benchmark; Furthermore, if Figure 5 As shown, step S2100 includes: Step S2110, traverse the performance fingerprint of each typical working condition in the first performance fingerprint library, extract the performance fingerprint of the first typical working condition, and extract the performance fingerprint of the first typical working condition. The predicted performance index of the typical working condition is calculated. The predicted performance index of a typical working condition is consistent with the preset Deviation from the performance evaluation benchmark for a typical operating condition; Step S2120, The deviation of all the prediction performance indicators of the typical working condition is weighted averaged to obtain the The comprehensive performance deviation score of typical working conditions; Step S2130, comparing the comprehensive performance deviation scores of all typical working conditions with a preset deviation threshold, and marking the typical working conditions that are higher than the deviation threshold as weak working conditions; Step S2140: Generate a first performance evaluation report.

[0062] Specifically, the first performance evaluation report is an assessment and diagnosis of the ideal performance of the unit under different typical operating conditions. By comparing the predicted performance in the first performance fingerprint library with industry standards, design indicators, etc., the deficiencies in the design and manufacturing of the unit can be discovered, providing a basis for subsequent optimization and improvement. This report is based on the inherent performance potential of the unit, so it is called the "first performance evaluation." The purpose of generating an evaluation report is to examine the performance characteristics of the unit from a macro perspective, identify the "shortcomings" that restrict its performance, and indicate the direction and space for optimization.

[0063] Load the first performance fingerprint library and obtain n 1 Typical working conditions. A kind of operating condition is selected, and various predicted performance indicators of the unit under this operating condition are extracted, such as power generation efficiency, fuel consumption rate, NOx emissions, etc. At the same time, according to the design documents of the unit, industry standards, etc., a performance evaluation benchmark under this typical operating condition is constructed. The evaluation benchmark gives the various performance indicators that the unit should achieve under this operating condition, representing the average or advanced level of similar units. Using appropriate similarity metrics (such as Euclidean distance, KL divergence, etc.), the deviation of each predicted indicator from the benchmark is quantitatively calculated. The greater the deviation, the greater the gap between the unit's performance on this indicator and the ideal level, and there is room for optimization. By summarizing the deviations of all predicted indicators, the overall deviation of the unit performance under this condition can be obtained.

[0064] Since different performance indicators have different effects on unit performance, it is necessary to set weights reasonably. Through expert scoring, hierarchical analysis and other methods, the weight coefficient of each indicator is determined to reflect its importance. Taking power generation efficiency and pollutant emissions as an example, in the context of focusing on energy conservation and emission reduction, both can be given greater weights. For peak-valley regulation power stations, the weights of indicators such as load change rate and start-stop time should be higher. The product of each deviation and weight is summed to obtain the comprehensive performance deviation of the operating condition, which is used as a quantitative score to evaluate the overall performance under the operating condition.

[0065] For 1 The comprehensive performance deviation scores of typical operating conditions are sorted. An appropriate deviation threshold is set (which can be given by expert experience), and typical operating conditions that are higher than the deviation threshold are marked as short-board conditions. Short-board conditions refer to conditions in which the performance of the unit deviates greatly from the ideal level, which restricts the improvement of the overall efficiency of the unit and urgently needs to be optimized. The selection of thresholds needs to balance sample coverage and optimization efforts. The higher the threshold, the fewer short-board conditions are, and the more targeted they are, but some conditions with optimization value may be missed. The lower the threshold, the more short-board conditions are, and the more comprehensive they are, but the optimization investment is large.

[0066] Based on the above analysis, the first performance evaluation report is generated, which mainly includes the following contents: (1) The ranking of comprehensive performance deviation under each typical operating condition and its comparison with the deviation threshold value; intuitively showing the performance of the unit under different operating conditions.

[0067] (2) A radar chart comparing the predicted performance indicators of the weak working condition with the benchmark, which details the degree of deviation of each indicator under the working condition and reveals the performance shortcomings.

[0068] (3) Optimization suggestions for weak operating conditions, such as equipment diagnosis and control logic upgrade, indicate the direction and measures for optimization.

[0069] (4) Statistical analysis of the overall performance level of the unit, such as average deviation, percentage of operating conditions that are better than the industry benchmark, etc., to quantitatively evaluate the performance potential of the unit.

[0070] The evaluation report uses a large number of charts to transform complex data into intuitive information. Through the report, designers and operation and maintenance personnel can comprehensively review the performance characteristics of the unit, focus on overcoming shortcomings, and maximize its efficiency potential. At the same time, the report can also serve as an important basis for communication with the owner, reflecting the differentiated advantages of the unit.

[0071] For example, for a certain type of gas turbine, its first performance fingerprint library contains three typical load conditions: 100%, 75%, and 50%. By comparing with the design indicators, it was found that the deviation of its comprehensive performance at 75% load was as high as 15%, exceeding the threshold of 10%, and was marked as a short board condition. Further analysis found that the compressor efficiency and turbine efficiency were seriously low under this condition, 5% and 7% lower than the design value respectively, which is the main short board. At the same time, it was noted that its deviation at 100% load was only 5%, which is better than the industry benchmark, reflecting the good peak capacity of the unit.

[0072] According to the evaluation results, the designers focused on analyzing the design parameters of the compressor and turbine components under 75% operating conditions, and compared them with similar models. Finally, they optimized the blade profile and clearance to make its efficiency close to the design value. The operation and maintenance personnel also adjusted the operation mode of this load point accordingly to avoid frequent operation in this area. Through optimization, the annual average power generation efficiency of the unit has increased by 1.2%, saving 5 million yuan in fuel costs each year.

[0073] It can be seen that the performance evaluation carried out according to the design conditions is an effective means to systematically diagnose the "congenital" deficiencies of the unit. Through the identification and quantitative analysis of the weak operating conditions, precise policies can be implemented to maximize the efficiency of the unit and realize its performance potential. This is the basic link of gas turbine performance management.

[0074] The above steps start from typical operating conditions, based on design and industry standards, quantitatively evaluate the performance of the unit under different operating conditions, identify the shortcomings that restrict its performance, and provide a handle for optimization. By loading the ideal performance of the first performance fingerprint library, the inherent potential of the unit is evaluated, which is the first level of performance evaluation.

[0075] Step S2200, generating a second performance evaluation report based on the first performance fingerprint library and the second performance fingerprint library; Furthermore, if Figure 6 As shown, step S2200 includes: Step S2210, traverse all historical performance fingerprints in the second performance fingerprint library, perform similarity matching with the first performance fingerprint library, and identify the operating condition of the unit; the operating condition of the unit is divided into typical operating condition and atypical operating condition; Further, step S2210 includes: Step S2211, calculate the first The historical performance fingerprint and the first performance fingerprint library Similarity of performance fingerprints , get the similarity sequence; 1≤ ≤n 2 , 1≤ ≤n 3 ;n 2 is the number of historical performance fingerprints in the second performance fingerprint library, n 3 is the number of performance fingerprints in the first performance fingerprint library, n 3 =n 1 , n 2 =n 4 ; The calculation of the second performance fingerprint library The historical performance fingerprint and the first performance fingerprint library Similarity of performance fingerprints include: in: The first item: weighted normalized Euclidean distance part (difference in working condition parameters) Meaning: Measures the normalized weighted distance between fingerprint parameters, capturing the differences in fingerprints in the main feature dimensions.

[0076] Explanation of symbols: : The total dimension of the first and second level typical operating parameters.

[0077] :The second performance fingerprint library The historical performance fingerprint parameter values.

[0078] :The first performance fingerprint library The performance fingerprint parameter values.

[0079] : No. The standard deviation of each parameter is used to normalize each parameter to eliminate the dimension effect. It measures the discreteness of the parameter. After normalization, it avoids the excessive influence of some features on the similarity calculation due to their large dimensions.

[0080] : No. The weight of a parameter reflects the importance of the parameter to the similarity calculation and is determined by feature selection methods (such as model-based feature importance and expert experience). The parameter weight can highlight the role of key performance indicators in similarity calculation.

[0081] Item 2: Nonlinear performance index difference Meaning: Measures the nonlinear difference in performance indicators between fingerprints.

[0082] Explanation of symbols: : The total number of performance indicators (such as power generation efficiency, fuel consumption rate, etc.).

[0083] : The second performance fingerprint library The historical performance fingerprint performance indicator value.

[0084] :The first performance fingerprint library The performance fingerprint performance indicator value.

[0085] : Nonlinear adjustment index, used to control the sensitivity of different performance indicators, set by historical data fitting or expert experience. Adjust the impact of the difference of each performance indicator on the similarity, the larger Emphasize the large differences between performance indicators.

[0086] Item 3: Comprehensive similarity formula (combining weights and nonlinear differences) Meaning: Reflect the distance between the first item and the second item as a similarity score , the distance is mapped to Similarity of range.

[0087] Explanation of symbols: : Balance coefficient, which controls the relative importance of the first term (difference in operating parameters) and the second term (difference in performance indicators) in similarity calculation. It is determined by parameter tuning, usually optimized on the training set through cross-validation. It balances the contribution of the two types of features to the similarity and avoids one type of feature dominating the calculation.

[0088] : An exponential function with the base e of the natural logarithm as its base.

[0089] Range is , as the difference in operating parameters increases (i.e. and The gap between them increases), the distance value of the first item increases, resulting in The index decreases, and the similarity decreases. As the difference in performance indicators increases (i.e. and The absolute difference between them increases), the nonlinear difference of the second term increases, resulting in a decrease in similarity. or nonlinear index As the difference in important features or performance indicators increases, it will have a greater impact on the similarity. The relative importance of control parameters and performance indicators. When increases, the difference of performance indicators has a greater impact on the similarity.

[0090] This formula takes into account the differences between operating parameters and performance indicators and constructs a global similarity metric. and nonlinear adjustment parameters , which can highlight the features and performance indicators that contribute most to similarity. By adjusting the balance coefficient and nonlinear index , which can adapt to the similarity requirements of different working conditions. Normalization eliminates the influence of dimension, so that different features can be compared on a unified scale. By combining weighted normalized Euclidean distance and nonlinear performance difference, the complex relationship between working condition parameters and performance indicators is fully characterized. Its complexity and nonlinear characteristics give it a significant advantage in capturing subtle differences between fingerprints and identifying anomalies. It can more accurately identify similar working conditions and potential abnormal working conditions, providing a solid foundation for subsequent performance evaluation and diagnosis.

[0091] Step S2212: If is the maximum value, and ≥ , then it is considered that The historical operating condition corresponding to the historical performance fingerprint belongs to Typical working conditions corresponding to each performance fingerprint; is the preset similarity threshold; Step S2213: If is the maximum value, and < , then the The historical operating conditions corresponding to the historical performance fingerprints are marked as atypical operating conditions.

[0092] Specifically, step S2210 aims to identify the actual operating conditions of the unit at historical moments through similarity matching. First, load the historical performance fingerprints from the second performance fingerprint library. Each historical performance fingerprint corresponds to the actual operating parameters and performance indicators of the unit at a certain historical moment, and comprehensively records the operating trajectory of the unit. Then, load the performance fingerprint in the first performance fingerprint library as a reference for condition identification.

[0093] "Similarity matching" is a common pattern recognition method. For the object to be identified (such as a historical performance fingerprint), the most similar pattern is searched in the known pattern library (such as the typical working condition performance fingerprint library) as its category. The similarity can be measured by the distance between vectors. The smaller the distance, the more similar it is. If the similarity between a historical performance fingerprint and all typical working condition fingerprints is less than , it is identified as a new atypical operating condition. It is a preset similarity threshold, usually with an empirical value such as 0.8. The emergence of atypical operating conditions indicates that the actual operating state of the unit deviates from the design operating conditions, and there may be performance problems such as equipment aging and parameter imbalance, which need to be paid attention to.

[0094] For example, it is assumed that the 100th historical performance fingerprint in the second performance fingerprint library is: {load:400MW,exhaustTemp: , steamPressure: 16MPa, efficiency: ; Among them, load is the load, exhaustTemp is the exhaust temperature, steamPressure is the main steam pressure, efficiency is the power generation efficiency, is the nitrogen oxide emission concentration.

[0095] Perform similarity matching with the performance fingerprint of the first performance fingerprint library to find the three most similar fingerprints: Typical working condition 1 (baseload): {load:450MW,exhaustTemp: , steamPressure: 16.5MPa, efficiency: , similarity: 0.9; Typical working condition 2 (75%_load): load:350MW,exhaustTemp: , steamPressure: 15MPa, efficiency: , similarity: 0.85; Typical working condition 4 (hot_day): {load:400MW,exhaustTemp: , steamPressure: 15.5MPa, efficiency: , similarity: 0.8; The result of the majority vote is typical operating condition 1, so it is determined that the unit is in full load (baseload) condition at that historical moment.

[0096] This operating condition identification method can reveal the degree of match between the actual operating status of the unit and the designed operating conditions. The frequent occurrence of atypical operating conditions means that the unit is difficult to adapt to the actual changes in grid demand. If the operating time is too short under some typical operating conditions, it reflects that the unit's performance in this condition is poor, and there may be local adaptability problems of the equipment. Based on the operating condition identification results, the utilization rate and change trend of the unit under each typical / atypical operating condition can be further statistically analyzed to find out the performance shortcomings, which will provide directions for subsequent abnormal diagnosis and performance optimization. Through the similarity matching algorithm, the historical operating data is linked to the ideal operating condition fingerprint, and an "actual-ideal" mapping can be constructed to characterize the unit's adaptability to the designed operating conditions using mathematical methods. The similarity calculation of the multi-dimensional parameter space can more accurately measure the unit status than a single performance indicator. Statistical analysis based on big data can reveal the relative advantages and disadvantages of individual units in the group.

[0097] Step S2210 can accurately identify the operating condition attribution of the historical performance fingerprint and quantify the gap between the actual operating condition and the designed operating condition. This lays the foundation for the subsequent evaluation of the unit performance from the perspective of actual operating capability. Through the operating condition label, the dynamic response of the unit performance to the change of operating conditions can be tracked, and the performance deterioration trend can be discovered in time. At the same time, it can also reveal the adaptation shortcomings of special operating conditions such as peak load regulation and deep peak load regulation, providing a handle for improving flexible performance.

[0098] Step S2220, based on the identification result of the operating condition of the unit, analyzing the cumulative operating time and start-stop frequency of the unit under each typical operating condition and atypical operating condition; Step S2230, based on the accumulated running time and start-stop frequency, the effective utilization rate of the computer group under each typical working condition and atypical working condition; Step S2240, analyzing T 3 The changing trend of the effective utilization rate of the unit under various atypical operating conditions within a time period; Step S2250, horizontally comparing the utilization differences of similar units under various historical operating conditions to find out the performance shortcomings of the units; Step S2260, generating a second performance evaluation report.

[0099] Specifically, step S2200 aims to generate a comprehensive performance evaluation report based on the first and second performance fingerprint libraries from the perspective of the long-term actual operating capacity of the unit. Based on historical data, the report quantitatively analyzes the utilization rate, start-stop characteristics, load adaptability, etc. of the unit under different operating conditions, and tracks the evolution trend of these characteristics vertically to identify performance shortcomings and make optimization suggestions. It complements the theoretical performance evaluation based on deviation in the first performance evaluation report, and together provides a basis for the full life cycle management of the unit.

[0100] Step S2210 identifies the operating conditions of the historical operating data in the second performance fingerprint library. Using algorithms such as similarity matching, each historical performance fingerprint is mapped to the most similar typical operating condition, or it is determined to be an atypical operating condition. This process essentially extracts valuable operating condition information from massive unlabeled operating data to clarify the actual working status of the unit. The frequent occurrence of atypical operating conditions means that the actual operation of the unit has seriously deviated from the design state, and there may be a risk of performance deterioration.

[0101] On this basis, steps S2220-S2230 count the cumulative operating time, start-stop times and other indicators of the unit according to the operating condition type, and quantitatively characterize the adaptability of the unit to different operating conditions; construct an indicator reflecting the effective utilization rate (Effective Utilization Rate, EUR) of the unit, which is the utilization rate of the unit under a certain operating condition, that is, the percentage of the actual operating time of the unit under a certain operating condition to the total operating time. Operating conditions with high utilization rates usually represent the main operating mode and areas of expertise of the unit. Operating conditions with high utilization rates usually represent the main operating mode and core advantages of the unit. Operating conditions with low utilization rates reflect the potential shortcomings of the unit, such as insufficient low-load peak-shaving capacity. Pay special attention to the utilization rate of the design conditions. If the utilization rate is too low, it means that the actual operation of the unit has seriously deviated from the design state.

[0102] Step S2240 further analyzes the change trend of the EUR of the unit under atypical operating conditions over time. 3The time series curve of EUR in the time period directly presents the dynamic adaptation process of the unit to non-design conditions. The monotonically rising curve indicates that the actual operating capacity of the unit under this condition is continuously improving, which may benefit from measures such as equipment transformation and operation optimization. The rapid decline of the curve warns that the adaptability of the unit under this condition is deteriorating, and the cause needs to be diagnosed in time. The curve shows obvious periodicity, which usually corresponds to the periodic changes in the external environment of the unit, such as seasonal fluctuations in grid demand.

[0103] Step S2250 performs a benchmark analysis on the EUR indicators of different units. Visualization methods such as radar charts are used to display the utilization differences between the target unit and similar units in key operating conditions in multiple dimensions. Find out the weak operating conditions (utilization rate is significantly lower than that of peers) and advantageous operating conditions (utilization rate is far ahead) of the target unit, and explore the reasons for the differences. The performance problems reflected by the weak operating conditions are often the focus of subsequent diagnosis and optimization.

[0104] Finally, step S2260 combines the above analysis to form a second performance evaluation report. The report systematically describes the actual long-term operating performance of the unit, focusing on topics such as quantitative evaluation of the unit's actual operating condition adaptability, historical change trend analysis, and performance benchmarking of similar units.

[0105] The second performance evaluation report mainly includes: (1) Statistical indicators such as the cumulative operating time, utilization rate, start-stop frequency, etc. of each historical operating condition, as well as the utilization rate change trend of atypical operating conditions in different time periods (such as monthly, quarterly, and annual). Through intuitive data statistics and trend analysis, managers can accurately grasp the actual long-term operating characteristics of the unit and gain insight into the laws of performance changes.

[0106] (2) Analysis of the causes of long-term changes in utilization under atypical conditions. In-depth analysis of key characteristic points in the utilization curve, such as inflection points, step increases / decreases, etc., combined with internal and external factors such as equipment status, grid dispatching requirements, and environmental conditions, to infer the deep-seated causes of performance changes. This is the basis for shifting from passive response to active management.

[0107] (3) Horizontally compare and analyze the utilization differences between the unit and similar units under key operating conditions, identify performance shortcomings, and make a preliminary diagnosis of the causes of the shortcomings. Visualize the results of the benchmarking analysis, and use radar charts, comparative bar charts, etc. to intuitively show the gap between the unit and its peers.

[0108] (4) For performance shortcomings, such as insufficient peak-shaving capacity and frequent starts and stops, macro-level optimization suggestions are proposed to provide direction for the formulation of specific diagnosis and treatment measures in the next step. Based on the design characteristics of the unit, historical experience, industry best practices, etc., possible solutions are preliminarily explored to gain the initiative for in-depth diagnosis.

[0109] The generation of the second performance evaluation report marks the successful transformation of the performance evaluation perspective, achieving a leap from the deviation analysis of static design to the dynamic actual capacity characterization. This idea breaks through the limitations of traditional thermal engineering calculations, and for the first time examines the unit in a long-term, dynamic production practice context, expanding the breadth and depth of performance evaluation from the two dimensions of time and space.

[0110] The performance shortcomings and historical change patterns revealed in the report are not only a summary of past operating practices, but also a guide to future optimization directions. The EUR indicator based on big data analysis is closer to the actual value of the unit than traditional KPIs (such as power generation efficiency, coal consumption rate, etc.), and is more easily accepted by owners and managers. The adaptability of the unit to operating conditions reflected by the dynamic change trend of EUR is precisely the core competitiveness that power companies are most concerned about in the context of flexible transformation.

[0111] In short, the second performance evaluation report echoes the first performance evaluation report, and three-dimensionally outlines the performance portrait of the unit at the two levels of theoretical design performance and actual operation performance. The combination of the two marks a revolutionary reconstruction of the unit performance evaluation system and is an important cornerstone for achieving full life cycle management. Standing at a new starting point, power companies will be able to diagnose performance problems more accurately, optimize production strategies more comprehensively, and adapt to the rapid evolution of grid demand more efficiently.

[0112] Step S2300, identifying abnormal operating conditions from typical operating conditions according to the first performance evaluation report and the second performance evaluation report, and generating an abnormal operating condition fingerprint library and an abnormal operating condition handling strategy library; Further, step S2300 includes: Step S2310: According to the comprehensive performance deviation score of the typical working condition, the typical working condition is divided into four abnormal levels: normal, slightly abnormal, moderately abnormal, and severely abnormal; the typical working conditions of slightly abnormal, moderately abnormal, and severely abnormal are marked as abnormal working conditions; and the normal typical working condition is marked as a stable working condition; Step S2320, extracting the performance indicator features of the abnormal operating condition from the performance indicators of the abnormal operating condition; Step S2330, generating an abnormal operating condition fingerprint library according to the operating condition parameters, performance indicator characteristics and abnormal level of the abnormal operating condition.

[0113] Step S2340, for abnormal operating conditions of different abnormal levels, configure abnormal operating condition handling strategies according to the first performance evaluation report and the second performance evaluation report, and generate an abnormal operating condition handling strategy library.

[0114] Specifically, the purpose of step S2300 is to further focus on abnormal operating conditions with large performance deviations on the basis of completing performance evaluation, conduct in-depth diagnosis, find out the causes of abnormalities, and lay the foundation for subsequent disposal. The core of abnormal diagnosis is to reveal the inherent laws and evolution trends of abnormal operating conditions through big data analysis and mechanism reasoning, and form an interpretable, predictable, and optimizable operating condition portrait.

[0115] In step S2310, based on the comprehensive performance deviation scores of each typical operating condition obtained from the first performance evaluation report, a cluster analysis method is used to classify the operating conditions into normal, slightly abnormal, moderately abnormal, and severely abnormal levels according to the degree of abnormality. Since there are many operating conditions of the gas-steam combined cycle unit, it is costly and difficult to carry out abnormal diagnosis one by one, so it is necessary to classify the degree of abnormality first so as to focus on investing diagnostic resources.

[0116] Specifically, a commonly used clustering algorithm such as K-means can be used to divide all working conditions into a preset number (such as 4) of abnormal levels by minimizing the sum of squares of deviations within the group. The comprehensive deviation score is used as the clustering feature to optimize the cluster center so that the working conditions in the same level have the highest consistency in the degree of abnormality.

[0117] This abnormality classification method can quickly screen out a few typical working conditions with high abnormality from a large number of working conditions, narrow the scope of subsequent abnormal diagnosis, and improve diagnostic efficiency. Clustering can reveal the similarities of different abnormal working conditions, providing a reference for summarizing the causes of abnormalities and formulating targeted abnormality handling strategies. Hierarchical management helps to differentiate the configuration of diagnostic resources, focus on and handle serious abnormalities, take into account minor abnormalities, and achieve a "cost-benefit" balance of abnormality management under limited resource conditions.

[0118] For example, a gas-steam combined cycle power plant identified a total of 100 typical operating conditions, and calculated that their comprehensive performance deviation scores were distributed between 0.5% and 12%. After K-means cluster analysis, 60 operating conditions with a deviation of less than 1% were classified as normal, 20 operating conditions with a deviation of 1% to 3% were classified as mild abnormalities, 15 operating conditions with a deviation of 3% to 6% were classified as moderate abnormalities, and 5 operating conditions with a deviation of more than 6% were classified as severe abnormalities. Focus on and deal with moderate and above abnormal conditions. In this way, subsequent abnormal diagnosis can be concentrated on 20 abnormal conditions, greatly improving resource utilization efficiency.

[0119] The performance indicator data of abnormal conditions comes from the first performance fingerprint database. From the performance indicator data of abnormal conditions, typical performance indicator features that reflect abnormal characteristics are extracted, such as low load, high coal consumption, high pollutant emissions, etc. These features help to intuitively describe the performance shortcomings of abnormal conditions and are the entry point for abnormal diagnosis and performance optimization. Performance indicator feature extraction can use data mining technologies such as statistical analysis and pattern recognition to find common laws of abnormal conditions in performance indicators. The operating parameters of abnormal conditions, the extracted performance indicator features, and the abnormal level are matched and integrated, and the key information is extracted and expressed and stored in a certain data structure (such as vectors, matrices, etc.) to generate an abnormal condition fingerprint database. The abnormal condition fingerprint database comprehensively depicts the performance characteristics of various abnormal conditions and is an important knowledge base for abnormal diagnosis, prediction, and optimization. The operating parameters of abnormal conditions come from the first performance fingerprint database and the typical operating condition parameter database.

[0120] For different abnormal levels of working conditions, match the preset abnormal handling strategies to form an abnormal working condition handling strategy library. Usually adopt the idea of ​​hierarchical management and tailor differentiated handling solutions: For severe abnormal operating conditions, it is necessary to formulate systematic optimization projects, carry out equipment technology transformation and management innovation, fundamentally eliminate abnormalities and improve unit performance levels.

[0121] For moderately abnormal operating conditions, they need to be listed as daily key focus areas, strengthen online monitoring and trend warnings, promptly detect and control performance deterioration, and prevent abnormalities from escalating.

[0122] Mild abnormal operating conditions can be included in the scope of routine management improvements, and performance deviations can be reduced and abnormal risks can be eliminated by optimizing operating processes, adjusting control parameters, and other means.

[0123] In addition, it is necessary to establish a long-term mechanism for graded handling of abnormal operating conditions, formulate judgment criteria for abnormal upgrades and releases, clarify the handling procedures, division of responsibilities, and assessment requirements for abnormal conditions at all levels, form a normalized abnormal management and control system, and continuously optimize abnormal handling capabilities.

[0124] After the above abnormal diagnosis steps, the managers of gas-steam combined cycle units can clearly understand the shortcomings of the current unit performance, the causes of abnormalities and their severity, so as to take appropriate measures according to the "symptoms" and deploy management resources in a targeted manner to achieve precise performance improvement. As a systematic knowledge base, the abnormal condition fingerprint library and disposal strategy library can guide the subsequent online monitoring, trend prediction and active optimization of unit performance, laying a solid foundation for full life cycle management.

[0125] Step S3000, perform online warning of abnormal operating conditions for the unit and output abnormal diagnosis results; according to the abnormal diagnosis results, match the abnormal operating condition handling strategy library to generate the unit's abnormal operating condition handling plan.

[0126] Furthermore, step S3000 includes: Step S3100, performing online warning of abnormal operating conditions on the unit and outputting abnormal diagnosis results; Furthermore, if Figure 7 As shown, step S3100 includes: Step S3110, collecting real-time operating parameters of the unit, and obtaining real-time predicted performance indicators according to the real-time operating parameters and the unit performance prediction model; Step S3120, generating a real-time performance fingerprint of the unit according to the real-time operating condition parameters and the real-time predicted performance index; Step S3130, perform similarity matching between the real-time performance fingerprint and the abnormal operating condition fingerprint library, and output the matching result.

[0127] Further, step S3130 includes: Step S3131, calculating the similarity between the real-time performance fingerprint and each abnormal operating condition fingerprint in the abnormal operating condition fingerprint library, and generating a second similarity sequence SI; Step S3132, find the maximum value from the second similarity sequence SI, recorded as SI A , A represents the abnormal condition fingerprint with the highest similarity to the real-time performance fingerprint in the abnormal condition fingerprint library; Step S3133: SI A and the preset abnormality judgment threshold θ 1 For comparison, if SI A <θ 1 , the current real-time operating condition is determined to be a stable operating condition; otherwise, it is determined to be an abnormal operating condition, and the abnormal level of the current real-time operating condition is determined according to the abnormal level of the abnormal operating condition corresponding to the abnormal operating condition fingerprint A.

[0128] Specifically, the purpose of step S3130 is to determine whether the current real-time operating condition is an abnormal operating condition, and if so, to determine its abnormality level. This step calculates the similarity between the real-time performance fingerprint and the samples in the abnormal operating condition fingerprint library to find the abnormal operating condition that is most similar to the current operating condition, and then determines whether it is abnormal and the degree of abnormality.

[0129] In step S3131, a similarity measurement method, such as Euclidean distance, cosine similarity, etc., is used to calculate the similarity between the real-time performance fingerprint and each fingerprint in the abnormal working condition fingerprint library to obtain a second similarity sequence SI. The greater the similarity (or the smaller the distance), the higher the similarity between the real-time working condition and the abnormal working condition, and the more likely it is to belong to the abnormal type.

[0130] For example, the Euclidean distances between the real-time performance fingerprint Pt and the three fingerprints P1, P2, and P3 in the abnormal operating condition fingerprint library are 2, 5, and 3 respectively, and the corresponding cosine similarities are 0.8, 0.5, and 0.7 respectively, then the second similarity sequence SI=(0.8, 0.5, 0.7).

[0131] Step S2132 selects the abnormal fingerprint with the greatest similarity from SI as the best match for the current real-time working condition. Continuing with the above example, the maximum similarity 0.8 is selected from SI=(0.8,0.5,0.7), and the corresponding abnormal working condition fingerprint is P1, so A=1. Step S3133 introduces the abnormal judgment threshold θ 1 , by comparing the maximum similarity SI A With θ 1 The size of SI can be used to determine whether the current working condition is abnormal. A <θ 1 , indicating that the current operating condition has a low similarity with any known abnormal operating condition and can be judged as a stable operating condition; otherwise, if SI A ≥θ 1 , then the similarity between the current working condition and the abnormal working condition fingerprint A exceeds the threshold, and it is determined to be an abnormal working condition, and the abnormal level is the same as the abnormal working condition of fingerprint A.

[0132] Abnormality judgment threshold θ 1 The setting of θ requires a trade-off between the sensitivity and false alarm rate of anomaly detection: 1 The larger the value, the lower the detection sensitivity, the higher the probability of missed detection, but the lower the false alarm rate; θ 1 The smaller it is, the higher the sensitivity is, and more anomalies can be found, but the false alarm rate increases. 1 It can be set based on experience or optimized through data analysis.

[0133] For example, the abnormality judgment threshold θ 1 is set to 0.75, the similarity SI between the current real-time performance fingerprint and the most similar abnormal condition (number A=1) A is 0.8, due to SI A =0.8>θ 1 =0.75, the current operating condition is judged to be abnormal, and the abnormality level is the same as abnormal condition 1. A =0.7<θ 1 =0.75, although the current operating condition is most similar to abnormal operating condition 1, the similarity is lower than the threshold and it is still judged as a stable operating condition.

[0134] In summary, step S3130 can quickly determine whether the current working condition is abnormal and what type of abnormality it belongs to by matching the similarity between the real-time performance fingerprint and the abnormal working condition fingerprint library, greatly improving the efficiency of real-time diagnosis of abnormal working conditions. This method makes full use of the knowledge base of the abnormal working condition fingerprint library constructed in step S2300, so that real-time abnormal diagnosis can be achieved through similarity matching instead of modeling and calculation, which has the advantages of simple calculation, timely response, and strong interpretability.

[0135] Step S3100 uses the operating parameters and performance prediction model collected in real time by the unit to dynamically generate a real-time fingerprint reflecting the current unit performance status, and achieves real-time warning and diagnosis of abnormal unit conditions by matching the similarity with the abnormal condition fingerprint library. Its purpose is to timely discover performance anomalies that occur during the operation of the unit, accurately locate the type and degree of anomalies, and provide a reliable basis for subsequent abnormal handling.

[0136] Step S3110 is the data preparation stage for generating real-time performance fingerprints. First, various real-time operating parameters are obtained through the data acquisition system of the unit, including load, ambient temperature, pressure, flow, etc. These parameters are the basis for evaluating the real-time performance of the unit. Among them, the load index reflects the power generation output level of the unit and is an important factor affecting other parameters; the ambient temperature affects the intake density and power output of the gas turbine; the combustion chamber pressure, main steam pressure, reheat steam pressure, etc. reflect the operating status of the unit cycle; the fuel flow, feed water flow, condensate flow, etc. reflect the material balance of the unit. In addition, it is necessary to collect the exhaust temperature, main steam temperature, etc. to determine the actual status of the unit equipment.

[0137] After obtaining the real-time operating parameters, they are input into the unit performance prediction model constructed in step S1200 to predict real-time performance indicators, such as power generation efficiency, heat rate, power supply coal consumption, etc. These indicators are key parameters for measuring the actual performance of the unit, but they are often not directly measurable and must be predicted with the help of models. The unit performance prediction model has been trained and optimized with historical operating data, and can extract deep performance-related features from complex operating parameters and establish a nonlinear mapping relationship between operating conditions and performance. Therefore, the model can accurately predict the performance of the unit under the current operating conditions based on the input real-time operating parameters. This model-based dynamic performance prediction can make up for the shortcomings of measured performance indicators and provide a more comprehensive and objective performance evaluation perspective.

[0138] Step S3120 is to extract features and fuse information of the acquired real-time operating parameters and predicted performance indicators, and finally form fingerprint features that reflect the real-time performance status of the unit. The key to this step is to design a reasonable fingerprint generation algorithm so that the generated real-time performance fingerprint can accurately and comprehensively describe the performance characteristics of the unit, which is convenient for subsequent similarity matching and abnormality judgment.

[0139] Real-time performance fingerprints are usually expressed in the form of vectors or matrices. Taking vector representation as an example, various operating parameters and performance indicators can be composed of several dimensions, and the value of each dimension corresponds to the real-time value of a parameter or indicator. Different parameters and indicators have different importance in judging unit performance, and necessary feature selection and weight assignment are required when constructing fingerprint vectors. Based on a combination of mechanism analysis and data-driven methods, the parameters that have the most significant impact on performance can be screened out, and the importance of different parameters can be measured using methods such as information gain and chi-square test, quantified into weight coefficients, and assigned to the corresponding dimensions of the fingerprint vector.

[0140] Since the dimensions and numerical ranges of different parameters and indicators vary greatly, in order to eliminate the impact of dimensions, the real-time data needs to be normalized before generating fingerprints. Common normalization methods include Min-Max normalization and Z-score normalization. Normalization maps the values ​​of each dimension to the same scale, making it easier to calculate the similarity between different fingerprints.

[0141] On the one hand, the real-time performance fingerprint must inherit the real-time characteristics of the operating parameters and be updated synchronously with the unit status; on the other hand, it must also take into account the comprehensive characteristics of the performance indicators and objectively reflect the inherent performance of the unit. The organic integration of the two enables the performance fingerprint to always record and characterize the changing trajectory of the unit performance in the dynamic evolution of the real-time operating conditions. Therefore, the real-time performance fingerprint can be regarded as a bridge to map and unify the external performance and internal mechanism of the unit, dynamically depicting the performance status of the unit "consistent inside and outside".

[0142] Step S3130 is to use mathematical methods to measure the closeness between the real-time performance fingerprint and the typical abnormal fingerprint in the abnormal condition fingerprint library, determine whether there is an abnormality in the current operating condition of the unit, and diagnose the type and severity of the abnormality. This step is the core of the entire abnormal warning and diagnosis process, and the scientificity and accuracy of its matching algorithm directly determines the effectiveness of warning and diagnosis.

[0143] The abnormal condition fingerprint library is a pre-built data set containing various known abnormal condition performance fingerprints. Each abnormal fingerprint corresponds to a specific abnormal condition, such as performance abnormalities caused by scaling, wear, component failure, etc. These fingerprints are derived from actual abnormal cases in the history of the unit, or from theoretical deductions based on mechanism analysis and experience summary by experts. The abnormal fingerprint and the real-time fingerprint use the same data structure and feature representation, so that the two are under the same judgment standard, which is convenient for similarity measurement.

[0144] The process of similarity matching is to calculate the "distance" between the real-time fingerprint and each fingerprint in the abnormal fingerprint library, and judge the similarity between the two by the size of the distance. The smaller the distance, the higher the similarity, and the closer the current working condition is to the abnormal working condition. The "distance" here can be measured by a variety of mathematical methods, the most common ones are Euclidean distance, Manhattan distance, Chebyshev distance, Mahalanobis distance, etc. Different distance metrics reflect different similarity criteria, such as Euclidean distance measures absolute numerical differences, Manhattan distance measures cumulative numerical differences, and the appropriate measurement method should be selected based on actual engineering experience. In the calculation process, the fingerprint is usually standardized to eliminate the influence of different feature dimensions.

[0145] The output of the matching results generally includes three aspects: one is the matching degree of the abnormal working condition, that is, the degree of proximity between the real-time fingerprint and the most similar abnormal fingerprint, which is usually expressed by a numerical value of 0~1. The closer to 1, the higher the matching degree; the second is the abnormality level, which divides the abnormality into mild, moderate, and severe according to the matching degree, reflecting the severity of the problem; the third is the most similar abnormal fingerprint, which provides a reference for diagnosing abnormalities.

[0146] In summary, step S3100 makes full use of the real-time data of the unit from multiple sources and heterogeneity, and combines it with the pre-established mechanism knowledge. Through the integration of data-driven and knowledge-driven, it dynamically constructs fingerprint features that can objectively reflect the real-time performance status of the unit. On this basis, intelligent algorithms such as similarity matching are used to compare and map the real-time operating conditions of the unit with typical abnormal operating conditions, and automatically realize abnormal warning and diagnosis. Compared with the traditional abnormal diagnosis method that relies on manual experience and threshold rules, this step can greatly improve the intelligence level and real-time performance of diagnosis, and can capture abnormal conditions more accurately, quickly and comprehensively, reducing omissions and misjudgments. At the same time, this method fully taps the value of historical data, optimizes the knowledge base with data-driven, guides practice with knowledge feedback, realizes the solidification and precipitation of empirical knowledge and real-time reuse, and escorts the safe and efficient operation of the unit.

[0147] Step S3200, according to the abnormal diagnosis result, match the abnormal condition handling strategy library and generate the abnormal condition handling plan of the unit.

[0148] Specifically, step S3200 automatically matches the pre-built abnormal operating condition handling strategy library based on the abnormal diagnosis result output by step S3100, forms a handling plan and control instructions for the current abnormal state of the unit, and sends them to the control system for execution, thereby realizing adaptive control and closed-loop management of the abnormal operating conditions of the unit.

[0149] The abnormal condition handling strategy library is a set of strategies that are pre-sorted and designed and correspond to the abnormal condition fingerprint library. Each type of abnormal fingerprint corresponds to several handling measures, which are usually extracted by experts based on mechanism analysis, simulation experiments, operating experience, etc., and represent the optimal control strategy under abnormal conditions. The handling measures include adjusting operating parameters, limiting load change rate, changing control methods, etc., to reduce the adverse effects of abnormal conditions on the unit and prevent abnormal deterioration.

[0150] The process of matching the abnormality handling strategy library is similar to the abnormality fingerprint matching principle of step S3130. It is to further clarify the control actions to be taken in response to the abnormality based on the diagnosis of the abnormality type. By calculating the correlation between the abnormality diagnosis results and the various handling measures in the strategy library, the control strategy that best matches the current abnormal state is selected. This "one-to-many" mapping mechanism can differentiate and treat the abnormalities according to their severity and urgency, so as to achieve accurate delivery of handling measures.

[0151] Based on the matching results, an abnormality handling plan is automatically generated, which mainly includes the following aspects: First, the diagnosed abnormality type, matching degree, abnormality level, etc., which comprehensively describe the current abnormal state of the unit; second, the control measures that should be taken, such as adjusting the setting values ​​of parameters such as fuel quantity, primary air volume, and water flow rate, changing the operation mode of the AGC (automatic generation control) system, etc., so that the unit can return to normal as soon as possible; third, the control instructions, usually a series of standard data frames, including control objects, operation types, values, time and other elements, can be directly sent to DCS and other control systems for execution.

[0152] The generation process of the disposal plan usually follows the principle of priority matching of abnormality levels. That is, according to the severity of the diagnosed abnormality, the disposal measures of the corresponding level are matched first. The abnormalities are divided into three levels: mild, moderate, and severe, corresponding to different disposal intensities and speeds. Mild abnormalities are mainly regulated on the spot, and the deviation is corrected by fine-tuning the operating parameters; moderate abnormalities require larger adjustments such as load limiting and cutting off auxiliary control; severe abnormalities require decisive implementation of safe shutdown. In addition, the coordination of different measures should be evaluated to avoid policy conflicts.

[0153] After the plan is generated, the control instructions are sent to the DCS system in a timely manner, which coordinates and controls the actuators of the combustion, water supply, steam turbine and other subsystems, and finally realizes adaptive adjustment of abnormal conditions. At the same time, the changes in the status of the unit are closely monitored, the treatment effect is evaluated, and the simulation model is called for post-evaluation when necessary to provide support for the dynamic optimization of the plan.

[0154] The matching and adaptive control of the abnormal handling strategy library have achieved a closed loop for the abnormal management of the unit. On the one hand, early warnings are triggered based on the diagnosis results, and on the other hand, adjustment instructions are quickly fed back. Through the seamless connection between diagnosis and decision-making, a real-time closed loop covering the entire process from abnormality generation to elimination is built. In particular, mild and moderate abnormalities can usually be eliminated through automatic adjustments, eliminating the burden of manual supervision. Even in the case of severe abnormalities, emergency response can be initiated at the first time to minimize accident losses.

[0155] It should be pointed out that although the solution can achieve a high degree of automation and intelligence, human interaction and decision-making links must still be embedded in key nodes. This is because unit anomalies are often accompanied by a large number of uncertainties, especially some extreme anomalies or compound anomalies, whose diagnosis and treatment require human understanding and judgment. Therefore, the best application mode of this solution is the deep integration of machine intelligence and expert experience to form a human-machine collaborative anomaly control paradigm. The idea of ​​machine recommendation and manual review can be used as a reference. The system provides a set of alternative disposal solutions, and experts make decisions based on actual conditions.

[0156] In general, step S3000 builds an effective abnormal condition warning and disposal scheme for the unit through a series of key technologies such as real-time fingerprint extraction, similarity matching, and strategy recommendation. The scheme is based on massive measured and simulated data, driven by mechanism models and expert experience, and integrates multiple intelligent technologies such as data analysis, fault diagnosis, and intelligent decision-making. It opens up the entire process of abnormal monitoring, analysis, early warning, diagnosis, and disposal, and realizes the end-to-end closed loop of unit abnormal management. This not only greatly improves the speed and accuracy of abnormal diagnosis and disposal, reduces accident losses, but also significantly reduces the intensity of manual duty and reduces the workload of operation and maintenance personnel. More importantly, while drawing on human experience and wisdom, the scheme also forms a new knowledge model through induction and deduction, realizing a virtuous cycle of human-computer interaction, knowledge precipitation and optimization. It can be foreseen that with the continuous accumulation of operating data and the continuous iteration of models, the performance of the scheme will be further improved, and it will surely become an important enabling technology for building smart power plants and realizing the full life cycle management of units.

[0157] Example 2 This embodiment provides a full life cycle management system for a gas-steam combined cycle unit based on the embodiment 1, such as Figure 8 As shown, including: Fingerprint library construction module: used to construct a typical operating condition parameter database and a unit performance prediction model, and to construct a first performance fingerprint library based on the typical operating condition parameter database and the unit performance prediction model; to obtain second historical operation data, and to construct a second performance fingerprint library based on the second historical operation data; Performance evaluation module: generates a first performance evaluation report based on the first performance fingerprint library and a preset performance evaluation benchmark; generates a second performance evaluation report based on the first performance fingerprint library and the second performance fingerprint library; Working condition diagnosis module: identifies abnormal working conditions according to the first performance evaluation report and the second performance evaluation report, and generates an abnormal working condition fingerprint library and an abnormal working condition disposal strategy library; Online analysis module: used to provide online warning of abnormal operating conditions of the unit and output abnormal diagnosis results; according to the abnormal diagnosis results, match the abnormal operating condition handling strategy library to generate the unit's abnormal operating condition handling plan.

[0158] In the fingerprint library construction module, the construction of the typical operating condition parameter database includes: Step S1110, obtaining gas-steam combined cycle unit n 1 The first level typical working condition parameters under the typical working condition; n 1 is the total number of typical working conditions; Step S1120, based on The first-level typical operating condition parameters under the typical operating condition are calculated to obtain the The second level typical working condition parameters under the typical working condition; 1≤ ≤n 1 ; Step S1130: The first level typical operating parameters and the second level typical operating parameters under the typical operating conditions The second level typical operating condition parameters under the typical operating condition constitute the Record the working parameters of typical working conditions; Step S1140: Record the operating parameters of all typical operating conditions to form a typical operating condition parameter database; the typical operating condition parameter database includes n 1 Group operating parameter records.

[0159] In the fingerprint library construction module, the construction of the unit performance prediction model includes: Step S1210, obtain the history T 1 The historical operation data of the gas-steam combined cycle unit within the time period is marked as the first historical operation data; the first historical operation data includes the first historical operating condition parameter and the first historical performance index; Step S1220: construct and train a unit performance prediction model based on the first historical operation data.

[0160] In the fingerprint library construction module, the construction of the first performance fingerprint library includes: Step S1310, according to typical operating parameters and unit performance prediction model, obtain n 1 Predictive performance indicators under typical working conditions; Step S1320, based on the typical operating condition parameter database The working parameter records of the typical working conditions and the The predicted performance index under typical working conditions is generated Performance fingerprint under typical working conditions; Step S1330: The performance fingerprints under all typical working conditions are combined into a first performance fingerprint library.

[0161] In the fingerprint library construction module, the construction of the second performance fingerprint library includes: Step S1410, obtaining history T 2 The historical operation data of the gas-steam combined cycle unit within the time period is marked as the second historical operation data; the second historical operation data includes the second historical operating condition parameter and the second historical performance index; the second historical operation data includes n 4 Group historical operating parameter records, n 4 is the total number of historical operating conditions; T 2 ≠T 1 ; Step S1420, generating a historical performance fingerprint based on the second historical operating condition parameter and the second historical performance index; Step S1430: All historical performance fingerprints are combined into a second performance fingerprint library.

[0162] In the performance evaluation module, generating the first performance evaluation report includes: Step S2110, traverse the performance fingerprint of each typical working condition in the first performance fingerprint library, extract the performance fingerprint of the first typical working condition, and extract the performance fingerprint of the first typical working condition. The predicted performance index of the typical working condition is calculated. The predicted performance index of a typical working condition is consistent with the preset Deviation from the performance evaluation benchmark for a typical operating condition; Step S2120, The deviation of all the prediction performance indicators of the typical working condition is weighted averaged to obtain the The comprehensive performance deviation score of typical working conditions; Step S2130, comparing the comprehensive performance deviation scores of all typical working conditions with a preset deviation threshold, and marking the typical working conditions that are higher than the deviation threshold as weak working conditions; Step S2140: Generate a first performance evaluation report.

[0163] In the performance evaluation module, generating the second performance evaluation report includes: Step S2210, traverse all historical performance fingerprints in the second performance fingerprint library, perform similarity matching with the first performance fingerprint library, and identify the operating condition of the unit; the operating condition of the unit is divided into typical operating condition and atypical operating condition; Step S2220, based on the identification result of the operating condition of the unit, analyzing the cumulative operating time and start-stop frequency of the unit under each typical operating condition and atypical operating condition; Step S2230, based on the accumulated running time and start-stop frequency, the effective utilization rate of the computer group under each typical working condition and atypical working condition; Step S2240, analyzing T 3 The changing trend of the effective utilization rate of the unit under various atypical operating conditions within a time period; Step S2250, horizontally comparing the utilization differences of similar units under various historical operating conditions to find out the performance shortcomings of the units; Step S2260, generating a second performance evaluation report.

[0164] The step S2210 includes: Step S2211, calculate the first The historical performance fingerprint and the first performance fingerprint library Similarity of performance fingerprints , get the similarity sequence; 1≤ ≤n 2 , 1≤ ≤n 3 ;n 2 is the number of historical performance fingerprints in the second performance fingerprint library, n 3 is the number of performance fingerprints in the first performance fingerprint library, n 3 =n 1 , n 2 =n 4 ; Step S2212: If is the maximum value, and ≥ , then it is considered that The historical operating condition corresponding to the historical performance fingerprint belongs to Typical working conditions corresponding to each performance fingerprint; is the preset similarity threshold; Step S2213: If is the maximum value, and < , then the The historical operating conditions corresponding to the historical performance fingerprints are marked as atypical operating conditions.

[0165] In the working condition diagnosis module, the identification of abnormal working conditions and the generation of abnormal working condition fingerprint library and abnormal working condition handling strategy library include: Step S2310: According to the comprehensive performance deviation score of the typical working condition, the typical working condition is divided into four abnormal levels: normal, slightly abnormal, moderately abnormal, and severely abnormal; the typical working conditions of slightly abnormal, moderately abnormal, and severely abnormal are marked as abnormal working conditions; and the normal typical working condition is marked as a stable working condition; Step S2320, extracting the performance indicator features of the abnormal operating condition from the performance indicators of the abnormal operating condition; Step S2330, generating an abnormal operating condition fingerprint library according to the operating condition parameters, performance indicator characteristics and abnormal level of the abnormal operating condition.

[0166] Step S2340, for abnormal operating conditions of different abnormal levels, configure abnormal operating condition handling strategies according to the first performance evaluation report and the second performance evaluation report, and generate an abnormal operating condition handling strategy library.

[0167] In the online analysis module, the abnormal operating condition online warning of the unit and the output of abnormal diagnosis results include: Step S3110, collecting real-time operating parameters of the unit, and obtaining real-time predicted performance indicators according to the real-time operating parameters and the unit performance prediction model; Step S3120, generating a real-time performance fingerprint of the unit according to the real-time operating condition parameters and the real-time predicted performance index; Step S3130, perform similarity matching between the real-time performance fingerprint and the abnormal operating condition fingerprint library, and output the matching result.

[0168] The step S3130 includes: Step S3131, calculating the similarity between the real-time performance fingerprint and each abnormal operating condition fingerprint in the abnormal operating condition fingerprint library, and generating a second similarity sequence SI; Step S3132, find the maximum value from the second similarity sequence SI, recorded as SI A , A represents the abnormal condition fingerprint with the highest similarity to the real-time performance fingerprint in the abnormal condition fingerprint library; Step S3133: SI A and the preset abnormality judgment threshold θ 1 For comparison, if SI A <θ 1 , the current real-time operating condition is determined to be a stable operating condition; otherwise, it is determined to be an abnormal operating condition, and the abnormal level of the current real-time operating condition is determined according to the abnormal level of the abnormal operating condition corresponding to the abnormal operating condition fingerprint A.

[0169] Example 3 This embodiment discloses an electronic device, which may include one or more processors and one or more memories, wherein the memories store computer-readable codes, and when the computer-readable codes are executed by the one or more processors, the full life cycle management method of the gas-steam combined cycle unit as described above may be executed.

[0170] The method or system according to the embodiment of the present application can also be implemented with the help of the architecture of an electronic device. The electronic device may include a bus, one or more CPUs, a read-only memory (ROM), a random access memory (RAM), a communication port connected to a network, an input / output component, a hard disk, etc. The storage device in the electronic device, such as a ROM or a hard disk, can store the full life cycle management method of the gas-steam combined cycle unit provided in the present application. The full life cycle management method of the gas-steam combined cycle unit may, for example, include: constructing a typical operating condition parameter database and a unit performance prediction model, and constructing a first performance fingerprint library based on the typical operating condition parameter database and the unit performance prediction model; obtaining the second historical operation data, and constructing a second performance fingerprint library based on the second historical operation data; generating a first performance evaluation report based on the first performance fingerprint library and a preset performance evaluation benchmark; generating a second performance evaluation report based on the first performance fingerprint library and the second performance fingerprint library; identifying abnormal operating conditions according to the first performance evaluation report and the second performance evaluation report, generating an abnormal operating condition fingerprint library and an abnormal operating condition handling strategy library; performing online warning of abnormal operating conditions on the unit, and outputting abnormal diagnosis results; matching the abnormal operating condition handling strategy library according to the abnormal diagnosis results, and generating an abnormal operating condition handling plan for the unit.

[0171] Furthermore, the electronic device may also include a user interface. Of course, the architecture disclosed in the present invention is only exemplary, and when implementing different devices, one or more components in the electronic device disclosed in the present invention may be omitted according to actual needs.

[0172] Example 4 This embodiment discloses a computer-readable storage medium, on which computer-readable instructions are stored. When the computer-readable instructions are executed by a processor, the full life cycle management method of the gas-steam combined cycle unit of the embodiment of the present application can be executed. The storage medium includes, but is not limited to, for example, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and cache memory (cache). Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc.

[0173] In addition, according to the implementation of the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the present application provides a non-transitory machine-readable storage medium, the non-transitory machine-readable storage medium stores machine-readable instructions, and the machine-readable instructions can be executed by a processor to execute instructions corresponding to the method steps provided in the present application, such as: constructing a typical operating condition parameter database and a unit performance prediction model, and constructing a first performance fingerprint library based on the typical operating condition parameter database and the unit performance prediction model; obtaining second historical operation data, and constructing a second performance fingerprint library based on the second historical operation data; generating a first performance evaluation report based on the first performance fingerprint library and a preset performance evaluation benchmark; generating a second performance evaluation report based on the first performance fingerprint library and the second performance fingerprint library; identifying abnormal operating conditions according to the first performance evaluation report and the second performance evaluation report, generating an abnormal operating condition fingerprint library and an abnormal operating condition handling strategy library; performing online warning of abnormal operating conditions on the unit, and outputting abnormal diagnosis results; matching the abnormal operating condition handling strategy library according to the abnormal diagnosis results, and generating an abnormal operating condition handling plan for the unit. When the computer program is executed by a central processing unit (CPU), the above functions defined in the method of the present application are executed.

[0174] The methods, systems, and devices of the present application may be implemented in many ways. For example, the methods, systems, and devices of the present application may be implemented by software, hardware, firmware, or any combination of software, hardware, and firmware. The above order of steps for the method is for illustration only, and the steps of the method of the present application are not limited to the order specifically described above unless otherwise specified. In addition, in some embodiments, the present application may also be implemented as programs recorded in a recording medium, which include machine-readable instructions for implementing the method according to the present application. Therefore, the present application also covers recording media storing programs for executing the method according to the present application.

[0175] In addition, the parts of the above-mentioned technical solutions provided in the embodiments of the present application that are consistent with the implementation principles of the corresponding technical solutions in the prior art are not described in detail to avoid excessive redundancy.

[0176] The specific implementation modes as described above further describe the purpose, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation mode of the present invention and is not intended to limit the present invention. 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 full life cycle management method for a gas-steam combined cycle unit, characterized in that: The method comprises: Constructing a typical operating parameter database and a unit performance prediction model, and constructing a first performance fingerprint library based on the typical operating parameter database and the unit performance prediction model; acquiring second historical operation data, and constructing a second performance fingerprint library based on the second historical operation data; Generate a first performance evaluation report based on the first performance fingerprint library and the preset performance evaluation benchmark; generate a second performance evaluation report based on the first performance fingerprint library and the second performance fingerprint library; identify abnormal working conditions according to the first performance evaluation report and the second performance evaluation report, and generate an abnormal working condition fingerprint library and an abnormal working condition handling strategy library; Provide online warning of abnormal operating conditions for the unit and output abnormal diagnosis results; match the abnormal operating condition handling strategy library based on the abnormal diagnosis results to generate the unit's abnormal operating condition handling plan.

2. The full life cycle management method of a gas-steam combined cycle unit according to claim 1, characterized in that: The construction of a typical operating condition parameter database comprises: Obtain the first-level typical operating condition parameters of the gas-steam combined cycle unit under n1 typical operating conditions; n1 is the total number of typical operating conditions; Based on The first-level typical operating condition parameters under the typical operating condition are calculated to obtain the The second level typical working condition parameters under the typical working condition; 1≤ ≤n1; The first The first level typical operating parameters and the second level typical operating parameters under the typical operating conditions The second level typical operating condition parameters under the typical operating condition constitute the Record the working parameters of typical working conditions; The operating condition parameter records of all typical operating conditions constitute a typical operating condition parameter database; the typical operating condition parameter database includes n1 groups of operating condition parameter records.

3. The full life cycle management method of a gas-steam combined cycle unit according to claim 2, characterized in that: The constructing of the unit performance prediction model comprises: Obtain historical operation data of the gas-steam combined cycle unit in the historical time period T1, marked as the first historical operation data; construct and train a unit performance prediction model based on the first historical operation data; The constructing of the first performance fingerprint library comprises: According to typical operating parameters and unit performance prediction model, the predicted performance indicators under n1 typical operating conditions are obtained; Based on the typical operating condition parameter database The working parameter records of the typical working conditions and the The predicted performance index under typical working conditions is generated Performance fingerprint under typical working conditions; The performance fingerprints under all typical working conditions constitute the first performance fingerprint library.

4. The full life cycle management method of a gas-steam combined cycle unit according to claim 3, characterized in that: The second historical operation data is the historical operation data of the gas-steam combined cycle unit in the historical time period T2, T2≠T1; the second historical operation data includes n4 groups of historical operating condition parameter records, n4 is the total number of historical operating conditions; The second historical operation data includes a second historical operating condition parameter and a second historical performance index; The constructing the second performance fingerprint library includes: generating a historical performance fingerprint based on a second historical operating condition parameter and a second historical performance index; All historical performance fingerprints are combined into a second performance fingerprint library.

5. The full life cycle management method of a gas-steam combined cycle unit according to claim 2, characterized in that: Generating the first performance evaluation report includes: Traverse the performance fingerprint of each typical working condition in the first performance fingerprint library and extract the The predicted performance index of the typical working condition is calculated. The predicted performance index of a typical working condition is consistent with the preset Deviation from the performance evaluation benchmark for a typical operating condition; For The deviation of all the prediction performance indicators of the typical working condition is weighted averaged to obtain the The comprehensive performance deviation score of typical working conditions; The comprehensive performance deviation scores of all typical working conditions are compared with a preset deviation threshold, typical working conditions that are higher than the deviation threshold are marked as weak working conditions, and a first performance evaluation report is generated.

6. The full life cycle management method of a gas-steam combined cycle unit according to claim 4, characterized in that: Generating the second performance evaluation report includes: Traverse all historical performance fingerprints in the second performance fingerprint library, perform similarity matching with the first performance fingerprint library, and identify the operating conditions of the unit; the operating conditions of the unit are divided into typical operating conditions and atypical operating conditions; Based on the identification results of the unit's operating conditions, analyze the unit's cumulative operating time and start-stop frequency under various typical and atypical operating conditions; Based on the cumulative operating time and start-stop frequency, the effective utilization rate of the computer group under various typical and atypical operating conditions is calculated; the changing trend of the effective utilization rate of the unit under various atypical operating conditions during the T3 time period is analyzed.

7. The full life cycle management method of a gas-steam combined cycle unit according to claim 6, characterized in that: The identification of the working condition of the unit includes: Calculate the second performance fingerprint library The historical performance fingerprint and the first performance fingerprint library Similarity of performance fingerprints , get the similarity sequence; 1≤ ≤n2,1≤ ≤n3; n2 is the number of historical performance fingerprints in the second performance fingerprint library, n3 is the number of performance fingerprints in the first performance fingerprint library, n3=n1, n2=n4; If the similarity sequence is the maximum value, and ≥ , then it is considered that The historical operating condition corresponding to the historical performance fingerprint belongs to Typical working conditions corresponding to each performance fingerprint; is the preset similarity threshold; If the similarity sequence is the maximum value, and < , then the The historical operating conditions corresponding to the historical performance fingerprints are marked as atypical operating conditions.

8. The full life cycle management method of a gas-steam combined cycle unit according to claim 5, characterized in that: The identifying of abnormal working conditions and generating an abnormal working condition fingerprint library and an abnormal working condition handling strategy library comprises: According to the comprehensive performance deviation score of typical working conditions, the typical working conditions are divided into four abnormal levels: normal, slightly abnormal, moderately abnormal, and severely abnormal; the typical working conditions with slightly abnormal, moderately abnormal, and severely abnormal are marked as abnormal working conditions; the normal typical working conditions are marked as stable working conditions; Extracting performance indicator characteristics of abnormal working conditions; generating an abnormal working condition fingerprint library according to working condition parameters, performance indicator characteristics and abnormal level of abnormal working conditions; For abnormal operating conditions of different abnormal levels, abnormal operating condition handling strategies are configured according to the first performance evaluation report and the second performance evaluation report, and an abnormal operating condition handling strategy library is generated.

9. The full life cycle management method of a gas-steam combined cycle unit according to claim 8, characterized in that: The online early warning of abnormal operating conditions of the unit includes: Collect the real-time operating parameters of the unit, and obtain the real-time predicted performance indicators based on the real-time operating parameters and the unit performance prediction model; Generate the real-time performance fingerprint of the unit based on the real-time operating parameters and real-time predicted performance indicators; Calculate the similarity between the real-time performance fingerprint and each abnormal working condition fingerprint in the abnormal working condition fingerprint library to generate a second similarity sequence SI; Find the maximum value from the second similarity sequence SI, recorded as SI A , A represents the abnormal condition fingerprint with the highest similarity to the real-time performance fingerprint in the abnormal condition fingerprint library; SI A Compared with the preset abnormality judgment threshold θ1, if SI A <θ1, the current real-time operating condition is determined to be a stable operating condition; otherwise, it is determined to be an abnormal operating condition, and the abnormal level of the current real-time operating condition is determined according to the abnormal level of the abnormal operating condition corresponding to the abnormal operating condition fingerprint A.

10. A full life cycle management system for a gas-steam combined cycle unit, which is used to implement a full life cycle management method for a gas-steam combined cycle unit according to any one of claims 1 to 9, characterized in that: The system comprises: Fingerprint library construction module: used to construct a typical operating condition parameter database and a unit performance prediction model, and to construct a first performance fingerprint library based on the typical operating condition parameter database and the unit performance prediction model; to obtain second historical operation data, and to construct a second performance fingerprint library based on the second historical operation data; Performance evaluation module: generates a first performance evaluation report based on the first performance fingerprint library and a preset performance evaluation benchmark; generates a second performance evaluation report based on the first performance fingerprint library and the second performance fingerprint library; Working condition diagnosis module: identifies abnormal working conditions according to the first performance evaluation report and the second performance evaluation report, and generates an abnormal working condition fingerprint library and an abnormal working condition disposal strategy library; Online analysis module: used to provide online warning of abnormal operating conditions of the unit and output abnormal diagnosis results; according to the abnormal diagnosis results, match the abnormal operating condition handling strategy library to generate the unit's abnormal operating condition handling plan.

Images