A steam turbine unit heat rate evaluation system and method based on historical data

Abstract

The present invention relates to the technical field of heat rate evaluation, and in particular to a heat rate evaluation system and method for a steam turbine unit based on historical data. The present invention generates a heat rate map by associating various heat rate-influencing parameters with historical heat rate data of the steam turbine unit, and initializes the associated weight of the heat rate map. Classification and marking are performed by analyzing the change trend of the heat rate, so as to analyze different heat rate change data; data analysis is performed on the classified heat rate data according to first heat rate influencing data and second heat rate influencing data through a heat rate parameter analysis model, so as to optimize the heat rate map weight to form a final heat rate map. By using this method, more accurate associated parameter data can be provided for analysis when evaluating the heat rate of the steam turbine unit, so as to reduce the workload of the heat rate evaluation work.

Description

Technical Field

[0001] The invention relates to the technical field of heat rate evaluation, and in particular to a system and method for evaluating the heat rate of a steam turbine unit based on historical data. Background Art

[0002] Heat rate is one of the important indicators for evaluating steam turbine units. The unit is usually kilojoules / kilowatt-hour, that is, the amount of heat consumed by the steam turbine unit for each kilowatt-hour of electricity generation. If the heat rate of the steam turbine unit is lower, it means that the less heat is required for power generation, and the thermal economy generated by the steam turbine unit is higher. Therefore, when evaluating steam turbine units, the lower the heat rate, the better.

[0003] The heat rate calculation of the steam turbine unit is affected by many factors. In addition to the hardware, such as equipment aging and improper maintenance will affect the heat rate; the performance parameters of the steam turbine unit, such as main steam temperature, exhaust vacuum, condenser performance parameters and other key parameters will affect the heat rate of the steam turbine unit; at the same time, during the maintenance process of the steam turbine unit, the various influencing parameters of the steam turbine unit will change, which will also affect the calculation of its heat rate.

[0004] Based on the above problems, the existing technology only takes corresponding solutions for a single problem, and collects and calculates a large amount of historical data during the operation of the steam turbine unit, but these data are not further processed and summarized. The current problem is that there are too many data that affect the heat rate calculation, which makes the heat rate evaluation work more complicated.

[0005] In order to further explore the potential rules of these historical data to simplify the heat rate evaluation work, the present invention proposes a steam turbine unit heat rate evaluation system and method based on historical data. Summary of the invention

[0006] The object of the present invention is to provide a system and method for evaluating heat rate of a steam turbine unit based on historical data, which generates a heat rate map by associating various parameters affecting the heat rate with the historical heat rate data of the steam turbine unit, and initializes the associated weights of the heat rate map, and classifies and marks the data by analyzing the changing trend of the heat rate, so as to analyze different heat rate change data; the classified heat rate data is analyzed according to the first heat rate influencing data and the second heat rate influencing data through a heat rate parameter analysis model, so as to optimize the heat rate map weights to form a final heat rate map, and through this method, more accurate associated parameter data can be provided for analysis when evaluating the heat rate of the steam turbine unit, so as to reduce the workload of the heat rate evaluation work.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] A steam turbine unit heat rate evaluation system based on historical data, comprising:

[0009] The historical heat rate acquisition module acquires the historical heat rate data of multiple steam turbine units;

[0010] a heat rate map establishing module, which associates the first heat rate influencing data and the second heat rate influencing data according to the historical heat rate data, generates a heat rate map, and initializes the association weight of the heat rate map;

[0011] A heat rate analysis module is used to classify the historical heat rate data of multiple groups of steam turbine units within a T period;

[0012] Further, according to the historical heat rate data of all the steam turbine units, the historical heat rate data is divided by taking the maintenance time T as a dividing point, and the data is classified according to the total number of the maintenance time T as the maintenance number;

[0013] generating a heat rate trend, comparing and marking the heat rate trend and calling the corresponding heat rate map;

[0014] Further, the heat rate trend of the historical heat rate data after segmentation is extracted by a linear regression algorithm, and the average of the heat rate trends with the same maintenance times is calculated as a standard heat rate trend;

[0015] Calculating the dispersion of the plurality of heat rate trends with the same number of maintenance times and the standard heat rate trend, dividing the trend into a plurality of levels according to the dispersion, marking the trends, and calling the heat rate graph of the corresponding category;

[0016] Comparing the heat rate trends of the same steam turbine unit at the nth maintenance times and the n+1th maintenance times according to the cosine similarity algorithm, if the comparison result is less than the heat rate trend threshold, calling the heat rate map of the steam turbine unit corresponding to the n+1th maintenance times;

[0017] a heat rate map first weight optimization module, which calls the first heat rate influencing data according to the heat rate map of the same category, extracts the first key feature and the second key feature of the first heat rate influencing data through a heat rate parameter analysis model, and optimizes the first weight of the heat rate map;

[0018] Furthermore, the first key feature and the second key feature are obtained according to an aging impact analysis model of the heat rate parameter analysis model; the aging impact analysis model includes an impact parameter preprocessing module, an impact parameter feature analysis module, a first weight optimization module and a first weight storage module;

[0019] The influencing parameter preprocessing module performs data segmentation and data standardization on the first heat rate influencing data to generate pre-input data; the influencing parameter feature analysis module predicts the data of the next segmented time period through 1 layer of Transformer and 6 layers of LSTM, and identifies the first key feature of the pre-input data and the second key feature of the data of the next segmented time period through 2 layers of residual convolution layers, 1 layer of dilated convolution layers and 4 layers of LSTM;

[0020] The first weight optimization module optimizes the association weight of the heat rate map according to the first key feature and the second key feature; specifically:

[0021] ;

[0022] in, is the first weight after optimization of the i-th first heat rate influencing data, is the initial association weight of the i-th first heat rate influencing data, The optimization parameters of the first key feature obtained by the aging impact analysis model, The optimization parameters of the second key feature obtained by the aging impact analysis model; is the key feature level label;

[0023] Further, The calculation process is:

[0024] ;

[0025] Among them, PD is the distance function, K is the k-means clustering algorithm, is the DTW algorithm, is the key feature of the input, Calculate key features for the rating, is the key feature discreteness, is the cluster center, is the cluster radius;

[0026] The first weight storage module stores the associated weights of the heat rate map;

[0027] a heat rate map second weight optimization module, which calls the second heat rate influencing data according to the heat rate map of the same category, extracts the third key feature and the fourth key feature of the second heat rate influencing data through the heat rate parameter analysis model, and optimizes the second weight of the heat rate map;

[0028] Furthermore, the third key feature and the fourth key feature are obtained according to the abnormality identification analysis model of the heat rate parameter analysis model; the abnormality identification analysis model includes an abnormal parameter preprocessing module, an abnormal parameter feature identification module, a second weight optimization module and a second weight storage module;

[0029] The abnormal parameter preprocessing module performs data segmentation and data standardization on the second heat rate impact data to generate pre-identification data; the abnormal parameter feature recognition module extracts the third key feature of the pre-identification data based on 1 layer of Transformer, 1 layer of time convolution layer and 4 layers of LSTM units, and extracts the fourth key feature of the pre-identification data through 2 layers of residual convolution layers, 1 layer of expansion convolution layer and 4 layers of LSTM;

[0030] The second weight optimization module optimizes the association weight of the second heat rate influencing data based on the third key feature and the fourth key feature; specifically:

[0031] ;

[0032] in, is the second weight after the j-th second heat rate influencing data is optimized, is the initial association weight of the jth second heat rate influencing data, is the optimization parameter of the third key feature obtained by the abnormality recognition analysis model, is the optimization parameter of the fourth key feature obtained by the anomaly recognition analysis model, and exp is an exponential function with the natural constant e as the base;

[0033] The second weight storage module stores the optimized association weight;

[0034] A heat rate map evaluation module is configured to call the first heat rate impact data and the second heat rate impact data corresponding to the heat rate map for evaluation during evaluation.

[0035] The present invention also proposes a method for evaluating the heat rate of a steam turbine unit based on historical data, specifically:

[0036] Acquire historical heat rate data of multiple groups of steam turbine units; associate first heat rate influencing data and second heat rate influencing data according to the historical heat rate data to generate a heat rate map, and initialize the association weight of the heat rate map;

[0037] Classifying the historical heat rate data of multiple groups of steam turbine units within a T period, generating heat rate trends, comparing and marking the heat rate trends and calling the corresponding heat rate graphs;

[0038] calling the first heat rate influencing data according to the heat rate map of the same category, extracting the first key feature and the second key feature of the first heat rate influencing data through a heat rate parameter analysis model, and optimizing the first weight of the heat rate map;

[0039] calling the second heat rate influencing data according to the heat rate map of the same category, extracting the third key feature and the fourth key feature of the second heat rate influencing data through the heat rate parameter analysis model, and optimizing the second weight of the heat rate map;

[0040] Furthermore, the heat rate parameter analysis model includes an aging impact analysis model and an abnormality recognition analysis model, including 2 Transformer layers, 18 LSTM layers, 2 dilated convolution layers, 4 residual convolution layers and 1 temporal convolution layer;

[0041] During the evaluation, the first heat rate impact data and the second heat rate impact data corresponding to the heat rate map are called for evaluation.

[0042] Compared with the prior art, the present invention has the following beneficial effects:

[0043] 1. The present invention classifies historical heat rate data of the same maintenance data and maintenance times, performs trend analysis of the heat rate data through a linear regression algorithm, takes the mean of the trend as the standard heat rate trend, divides the data into multiple levels according to the discreteness of the data, and calls the corresponding heat rate map according to the trend similarity result calculated by the cosine similarity algorithm; this method can classify and analyze the data according to the historical heat rate data from the longitudinal data dimension of different steam turbine units and the lateral data dimension of the same steam turbine unit, and further optimize the weight of the heat rate map based on the classified data results, so as to perform more accurate analysis based on the heat rate data under different situations, so as to provide more accurate heat rate influencing parameters under different situations.

[0044] 2. The present invention uses an aging impact analysis model of a heat rate analysis model to call all influencing parameters based on historical heat rate data of steam turbine units that have been maintained for multiple times. It can analyze the change characteristics of each influencing parameter from multiple time dimensions through global trend change analysis of current data and global trend change analysis of predicted data, so as to calculate the weight of each influencing parameter and optimize the heat rate map, provide more accurate related parameters for the historical heat rate data of aging steam turbine units, and facilitate reducing the evaluation and analysis work of the heat rate data of aging steam turbine units.

[0045] 3. The present invention uses an abnormal recognition analysis model of a heat rate analysis model to perform data analysis of abnormal influencing parameters on both new and old steam turbine units. The number of abnormal influencing parameter data and the trend of abnormal influencing parameter data of each influencing parameter are used to analyze the influence of each influencing parameter on the heat rate data when an abnormal influencing parameter occurs in the steam turbine unit from the dimension of data abnormality characteristics. This method can extract influencing parameters that have a greater impact on the heat rate and are prone to abnormalities, so as to reduce the evaluation and analysis work of the heat rate data of these steam turbine units. At the same time, these influencing parameters that are prone to abnormalities can also be extracted to facilitate adjustment and maintenance by relevant technical personnel. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 is a flow chart of a virtual system of the present invention;

[0047] Figure 2 It is a virtual device corresponding diagram of the heat rate parameter analysis model of the present invention;

[0048] Figure 3 A network structure diagram of an aging influence analysis model of a heat rate parameter analysis model of the present invention;

[0049] Figure 4 : is a network structure diagram of the residual convolution layer of the present invention;

[0050] Figure 5 A network structure diagram of an abnormality identification analysis model of a heat rate parameter analysis model of the present invention;

[0051] Figure 6 It is a calculation result diagram of DTW of the present invention;

[0052] Figure 7 is a flow chart of the method of the present invention;

[0053] Figure 8 A loss function diagram for the third key feature identification of the anomaly identification and analysis model of the present invention;

[0054] Fig. 9 This is a loss function diagram for the third key feature recognition of the Transformer+4-layer LSTM of the present invention. DETAILED DESCRIPTION

[0055] 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.

[0056] The heat rate evaluation of a steam turbine unit is one of the important economic indicators for a steam turbine unit. Since the evaluation and calculation of the heat rate involves a large number of related parameters, and parameter adjustment is easy to occur during the maintenance process, resulting in changes in parameter data affecting the heat rate evaluation, making the heat rate evaluation work more complicated. Therefore, the present invention provides a steam turbine unit heat rate evaluation system and method based on historical data, referring to Figure 1 As shown, the technical solution is as follows:

[0057] The historical heat rate acquisition module acquires the historical heat rate data of multiple steam turbine units;

[0058] a heat rate map establishing module, which associates the first heat rate influencing data and the second heat rate influencing data according to the historical heat rate data, generates a heat rate map, and initializes the association weight of the heat rate map;

[0059] A heat rate analysis module, which classifies the historical heat rate data of multiple groups of steam turbine units within a T period, generates heat rate trends, compares and marks the heat rate trends, and calls the corresponding heat rate graphs;

[0060] a heat rate map first weight optimization module, which calls the first heat rate influencing data according to the heat rate map of the same category, extracts the first key feature and the second key feature of the first heat rate influencing data through a heat rate parameter analysis model, and optimizes the first weight of the heat rate map;

[0061] a heat rate map second weight optimization module, which calls the second heat rate influencing data according to the heat rate map of the same category, extracts the third key feature and the fourth key feature of the second heat rate influencing data through the heat rate parameter analysis model, and optimizes the second weight of the heat rate map;

[0062] A heat rate map evaluation module is configured to call the first heat rate impact data and the second heat rate impact data corresponding to the heat rate map for evaluation during evaluation.

[0063] Through multiple groups of historical heat rate data of steam turbine units, a heat rate map is generated based on the historical heat rate data and all influencing parameters that may affect the heat rate, and the associated weights of each influencing data in the map are initialized to establish an initial relationship network; the maintenance time of the steam turbine unit is used as a cycle to obtain the corresponding maintenance times to classify the existing historical heat rate data and improve the accuracy of the data during analysis; trend analysis is performed on the classified historical heat rate data to mark the data so as to form corresponding heat rate maps for different problems of the heat rate data for subsequent use;

[0064] Based on calling the heat rate map, the influencing parameters of the aging steam turbine unit are used as the first heat rate influencing data, and the influencing parameters of the steam turbine unit with abnormal data are used as the second heat rate influencing data; prediction and trend analysis are used as the breakthrough point of the aging steam turbine unit, and the results of current data and predicted data are analyzed to realize multi-dimensional data feature extraction to optimize related parameters; abnormal trends and abnormal quantities are used as the breakthrough points of the steam turbine unit with abnormal data, and the specific abnormal data characteristics of each influencing parameter are analyzed to obtain the key related influencing parameters that cause fluctuations in the heat rate data to optimize the corresponding weights;

[0065] Based on the optimized heat rate map, the corresponding parameters can be obtained for evaluation based on different types of heat rate maps under different requirements, reducing the workload of large-scale parameter processing and calculation in the current heat rate evaluation work, improving the efficiency of the heat rate evaluation work, and also providing data reference for the maintenance work of relevant technical personnel to reduce the workload of maintenance work. At the same time, it can also facilitate the traceability of heat rate evaluation work after maintenance, further reducing the workload of heat rate evaluation work.

[0066] Embodiment 1

[0067] For the purpose of specific explanation, the following contents are combined for elaboration:

[0068] A steam turbine unit heat rate evaluation system based on historical data, comprising:

[0069] The historical heat rate acquisition module acquires the historical heat rate data of multiple steam turbine units;

[0070] A heat rate map establishment module is configured to associate the first heat rate influencing data and the second heat rate influencing data according to the historical heat rate data, generate a heat rate map, and initialize the association weight of the heat rate map; the default association weight is 1; the first heat rate influencing data is the influencing parameter of the steam turbine unit with aging equipment, and the second heat rate influencing data is the influencing parameter of the steam turbine unit with abnormal operation; the influencing parameters are, for example, but not limited to, steam temperature, pressure, condenser vacuum, etc.;

[0071] A heat rate analysis module is used to classify the historical heat rate data of multiple groups of steam turbine units within a T period (for example, but not limited to 1 month, 2 months, 3 months, etc.);

[0072] Further, according to the historical heat rate data of all the steam turbine units, the historical heat rate data is divided by taking the maintenance time T as a dividing point, and the data is classified according to the total number of the maintenance time T as the maintenance number; for example, the size of T is set to 1 month, the historical heat rate data within the first T time is classified, the historical heat rate data within the second T time is classified, the historical heat rate data within the third T time is classified, and so on;

[0073] With the increase of the service life and maintenance times of steam turbine units, the historical heat rate trend law of steam turbine units under different service times is different. Therefore, based on the maintenance time T as a reference indicator, multiple maintenance times T are used as the times for division, so as to analyze the steam turbine units under different situations and improve the accuracy of the analysis, so as to optimize the weight of the subsequent heat rate map related data and provide a reliable data basis for the evaluation of heat rate data;

[0074] generating a heat rate trend, comparing and marking the heat rate trend and calling the corresponding heat rate map;

[0075] Further, the heat rate trend of the historical heat rate data after segmentation is extracted by a linear regression algorithm (for example, but not limited to the least squares method, ridge regression, Lasso regression, etc.), and the average of the heat rate trend with the same maintenance times is calculated as the standard heat rate trend; for example, the historical heat rate data within the 5th T time after classification is subjected to a linear regression algorithm. In this embodiment, the least squares method is used to obtain multiple slopes of the fitting line after the linear regression algorithm. , n is the total number of historical heat rate data within the 5th T time, each data corresponds to a group of steam turbine units, and the calculation The mean slope k of is taken as the standard heat rate trend, and the corresponding fitting straight line is associated with the standard heat rate trend;

[0076] Calculate the dispersion of the multiple heat rate trends and the standard heat rate trend for the same number of maintenance times, divide the trend into multiple levels according to the dispersion, mark them, and call the heat rate map of the corresponding category; for example, the standard heat rate trend of the historical heat rate data within the 5th T time is a fitting straight line with a slope k, calculate the minimum vertical distance between each heat rate data within the 5th T time and the fitting straight line of the standard heat rate trend as the dispersion, and sum all the dispersion data to obtain the corresponding trend data;

[0077] The trend level is obtained based on the experience of relevant technical personnel and data statistics results, generating multiple discrete threshold intervals, setting a corresponding trend level label for each discrete threshold, and obtaining the corresponding trend level based on which discrete threshold interval the trend data is in;

[0078] Comparing the heat rate trends of the same steam turbine unit at the nth maintenance times and the n+1th maintenance times according to the cosine similarity algorithm, if the comparison result is less than the heat rate trend threshold, calling the heat rate map of the steam turbine unit corresponding to the n+1th maintenance times; the heat rate trend threshold is set according to expert experience;

[0079] The changing rules of historical heat rate data of steam turbine units under different service time conditions may be different. Therefore, the historical heat rate data are fitted by linear regression algorithm. The slope of the sum is fitted to understand the changing trend characteristics of the historical heat rate data. The reference trend data under different maintenance times and maintenance times are obtained based on the mean calculation, so that the data discreteness is calculated according to the data and divided into different levels. The data type analysis is further carried out in detail from the vertical data dimension.

[0080] At the same time, according to the change law of the data itself, the change law of the historical heat rate data of the steam turbine unit with the increase of maintenance times and usage time is analyzed from the horizontal data dimension; through the above method, a more comprehensive data dimension analysis is achieved, so as to divide a variety of heat rate maps for subsequent call and analysis, reduce the workload of evaluation and improve the accuracy of analysis;

[0081] The first weight optimization module of the heat rate spectrum calls the first heat rate influencing data according to the heat rate spectrum of the same category, extracts the first key feature and the second key feature of the first heat rate influencing data through the heat rate parameter analysis model, and optimizes the first weight of the heat rate spectrum; in the process of data input, mainly screens and inputs the influencing parameter data corresponding to the n+1th maintenance times of the steam turbine units whose cosine similarity calculation results are less than the heat rate trend threshold and which are marked as aging equipment by relevant technicians; these influencing parameter data also correspond to the historical heat rate data with different trend level labels;

[0082] Further, the first key feature and the second key feature are obtained according to the aging impact analysis model of the heat rate parameter analysis model; referring to Figure 2 As shown, the aging impact analysis model includes an impact parameter preprocessing module, an impact parameter feature analysis module, a first weight optimization module and a first weight storage module; the first weight is the weight of the relationship between the impact parameters of the heat rate of the aging steam turbine unit;

[0083] The influencing parameter preprocessing module performs data segmentation and data standardization on the first heat rate influencing data to generate pre-input data; data segmentation is to segment the data by setting the data window length according to expert suggestions, and data standardization is performed by using Z-score;

[0084] Reference Figure 3 As shown, the influencing parameter feature analysis module predicts the data of the next segmentation time period through 1 layer of Transformer and 6 layers of LSTM, and identifies the first key feature of the pre-input data and the second key feature of the data of the next segmentation time period through 2 layers of residual convolution layers, 1 layer of dilated convolution layers and 4 layers of LSTM; Transformer corresponds to Figure 3 The converter model in LSTM corresponds to Figure 3 The network structure of the residual convolution layer in the long short-term memory neural network is referenced Figure 4 As shown; the initial size of the dilated convolutional layer is , the parameter of each expansion interval is set to 1, and the expansion is performed 3 times;

[0085] The first key feature is the abnormal trend feature label of each influencing parameter’s current data, and the second key feature is the abnormal trend feature label of each influencing parameter’s predicted data; trend feature labels are mainly divided into two categories, normal trend feature labels and abnormal trend feature labels, the normal trend feature label is 0, and the abnormal trend feature label is 1;

[0086] Through the aging impact analysis model, further data analysis can be performed based on the aging steam turbine units that have been used for a long time. By analyzing the global status of each influencing data in the prediction results and the current historical data results, the key data affecting its heat rate can be grasped, and its data weight can be enhanced to facilitate the subsequent use of the heat rate map, thereby reducing the data evaluation and analysis work for the aging steam turbine units.

[0087] The first weight optimization module optimizes the association weight of the heat rate map according to the first key feature and the second key feature; specifically:

[0088] ;

[0089] in, is the first weight after optimization of the i-th first heat rate influencing data, is the initial association weight of the i-th first heat rate influencing data, The optimization parameters of the first key feature obtained by the aging impact analysis model, The optimization parameters of the second key feature obtained by the aging impact analysis model; is the key feature level label;

[0090] Based on the trend characteristics of each parameter data influencing the analyzed current data, the current data influencing the parameter data is used as the main parameter, and the predicted data influencing the parameter data is used as the auxiliary parameter. Through this method, the relationship between the parameter data and the heat rate data can be comprehensively analyzed from multiple time dimensions, so as to optimize its first weight to achieve a more accurate heat rate map, facilitate the relevant parameter call of the subsequent evaluation work, and reduce the workload of the evaluation work;

[0091] Further, The calculation process is:

[0092] ;

[0093] Among them, PD is the distance function, K is the k-means clustering algorithm, It is the DTW algorithm (Dynamic TimeWarping, dynamic time warping algorithm), is the key feature of the input, Calculate key characteristics for the rating, is the key feature discreteness, is the cluster center, is the clustering radius; the key feature is the number of different data calculated by the DTW algorithm, Reference parameter data set for experts;

[0094] The specific clustering process of the K-means algorithm is to use the number of different DTW data as the horizontal axis, refer to Figure 6 As shown, Figure 6 The gray part is the characteristic position when the curve data is consistent, and the black part is the actual characteristic position. Figure 6 The number of black parts in the gray part; the discreteness data is used as the ordinate to generate a two-dimensional coordinate system for clustering. The number of k is set based on expert experience. The discreteness data is calculated as follows:

[0095] ;

[0096] in, is the standard deviation formula, is the mean formula, for The corresponding data includes the i-th first heat rate impact data and the j-th second heat rate impact data;

[0097] In order to quantify the trend change data, the DTW algorithm is used to compare the different characteristics of each influencing parameter and the standard influencing parameter as the first coordinate, and the data dispersion of each influencing parameter is used as the second coordinate. The cluster center is obtained by clustering according to the k-means algorithm, and then the cluster radius is combined to realize the quantification of the trend data, generate the corresponding parameter data to optimize the first weight data, and realize more accurate heat rate map data;

[0098] The first weight storage module stores the associated weights of the heat rate map;

[0099] a heat rate spectrum second weight optimization module, calling the second heat rate influencing data according to the heat rate spectrum of the same category, extracting the third key feature and the fourth key feature of the second heat rate influencing data through the heat rate parameter analysis model, and optimizing the second weight of the heat rate spectrum; the input data includes the influencing parameter data of the aged steam turbine unit and the influencing parameter data of the non-aged steam turbine unit, and also includes the parameter data whose calculation result of the cosine similarity algorithm is less than the heat rate trend threshold and the influencing parameter data of the heat rate data of different levels;

[0100] Furthermore, the third key feature and the fourth key feature are obtained according to the abnormality identification analysis model of the heat rate parameter analysis model; the abnormality identification analysis model includes an abnormal parameter preprocessing module, an abnormal parameter feature identification module, a second weight optimization module and a second weight storage module;

[0101] The abnormal parameter preprocessing module performs data segmentation and data standardization on the second heat rate impact data to generate pre-identification data; referring to Figure 5 As shown, the abnormal parameter feature recognition module extracts the third key feature of the pre-identified data based on 1 layer of Transformer, 1 layer of temporal convolution layer and 4 layers of LSTM units, and extracts the fourth key feature of the pre-identified data through 2 layers of residual convolution layer, 1 layer of dilated convolution layer and 4 layers of LSTM; the third key feature is the number of abnormal features of each influencing parameter, and the fourth key feature is the abnormal feature trend of each influencing parameter; Transformer corresponds to Figure 5 The converter in LSTM corresponds to Figure 5 Long short-term memory neural networks in ;

[0102] The abnormal identification analysis model of the heat rate parameter analysis model can analyze the data in all use time periods and maintenance time periods. By mastering the number of abnormal change feature data and abnormal change trends of each influencing parameter during use, the connection between the influencing parameter and the heat rate data is enhanced, and the weight of the heat rate map is increased to facilitate the use of subsequent evaluation work and reduce the workload of heat rate evaluation work;

[0103] The second weight optimization module optimizes the association weight of the second heat rate influencing data based on the third key feature and the fourth key feature; specifically:

[0104] ;

[0105] in, is the second weight after the j-th second heat rate influencing data is optimized, is the initial association weight of the jth second heat rate influencing data, is the optimization parameter of the third key feature obtained by the abnormality recognition analysis model, is the optimization parameter of the fourth key feature obtained by the abnormal identification and analysis model, and exp is an exponential function with the natural constant e as the base; in the process of abnormal analysis, the number of abnormal feature data identified is used as the main parameter, and the abnormal change trend is used as the auxiliary parameter. Through this method, the abnormal change influencing parameters affecting the heat rate data can be grasped to the greatest extent, so as to facilitate the data quantification of relevant parameters of its abnormal change characteristics, optimize the initialized second weight data from the perspective of abnormal data, realize a more accurate heat rate map, and facilitate the generation and call of the heat rate map;

[0106] The second weight storage module stores the optimized association weight;

[0107] A heat rate map evaluation module is configured to call the first heat rate impact data and the second heat rate impact data corresponding to the heat rate map for evaluation during evaluation.

[0108] Embodiment 2

[0109] The present invention also proposes a method for evaluating the heat consumption rate of a steam turbine unit based on historical data, referring to Figure 7 As shown, specifically:

[0110] Acquire historical heat rate data of multiple groups of steam turbine units; associate first heat rate influencing data and second heat rate influencing data according to the historical heat rate data to generate a heat rate map, and initialize the association weight of the heat rate map;

[0111] Classifying the historical heat rate data of multiple groups of steam turbine units within a T period, generating heat rate trends, comparing and marking the heat rate trends and calling the corresponding heat rate graphs;

[0112] calling the first heat rate influencing data according to the heat rate map of the same category, extracting the first key feature and the second key feature of the first heat rate influencing data through a heat rate parameter analysis model, and optimizing the first weight of the heat rate map;

[0113] calling the second heat rate influencing data according to the heat rate map of the same category, extracting the third key feature and the fourth key feature of the second heat rate influencing data through the heat rate parameter analysis model, and optimizing the second weight of the heat rate map;

[0114] Furthermore, the heat rate parameter analysis model includes an aging impact analysis model and an abnormality recognition analysis model, including 2 Transformer layers, 18 LSTM layers, 2 dilated convolution layers, 4 residual convolution layers and 1 temporal convolution layer;

[0115] In this embodiment, the present invention compares the accuracy of the first key feature predicted by the aging impact analysis model and the third key feature identified by the abnormality recognition analysis model during the training process, and takes the highest accuracy value of each model for comparison, as shown in Table 1 and Table 2:

[0116] Table 1 Comparison results of the prediction accuracy of the first key feature

[0117]

[0118] As can be seen from Table 1, the prediction accuracy of the model in this paper is 90.61%, which is a high prediction accuracy;

[0119] Table 2 Comparison results of the third key feature recognition accuracy

[0120]

[0121] From Table 2, we can see that Transformer+4-layer LSTM has higher recognition accuracy than the model in this paper, but the change in loss function is as follows Figure 8 and Fig. 9 As shown in Figure 2, both models were iterated 140 times. Figure 8 The loss function in the training set and the test set converges, and the effect is good, while the Transformer+4-layer LSTM Fig. 9 The training set converged but the test set overfitted, so the effect was poor;

[0122] In this embodiment, the accuracy of the second key feature identified by the aging impact analysis model and the fourth key feature identified by the abnormality recognition analysis model during the training process is also compared. Similarly, the highest accuracy value of each model during the training process is taken for comparison. The specific results are shown in Table 3:

[0123] Table 3 Comparison of recognition accuracy of the second key feature and the fourth key feature

[0124]

[0125] In the training process, the 2-layer residual convolution layer, 1-layer dilated convolution layer and 4-layer LSTM of the proposed model are trained with the same model structure and different training data. As can be seen from Table 3, the recognition accuracy of the proposed model in the second key feature and the fourth key feature is higher;

[0126] During the evaluation, the first heat rate impact data and the second heat rate impact data corresponding to the heat rate map are called for evaluation.

[0127] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A steam turbine unit heat rate evaluation system based on historical data, characterized in that: include: The historical heat rate acquisition module acquires the historical heat rate data of multiple steam turbine units; A heat rate map establishment module is used to associate the first heat rate influencing data and the second heat rate influencing data according to the historical heat rate data, generate a heat rate map, and initialize the association weight of the heat rate map; The first heat rate influencing data is the influencing parameters of the steam turbine unit with aging equipment, and the second heat rate influencing data is the influencing parameters of the steam turbine unit with abnormal operation; the influencing parameters include: steam temperature, pressure and condenser vacuum; The heat rate analysis module classifies the historical heat rate data of multiple steam turbine units within the T period, generates heat rate trends, compares and marks the heat rate trends, and calls the corresponding heat rate maps; The process of comparing the heat rate trend and calling the corresponding heat rate map includes: extracting the heat rate trend of the segmented historical heat rate data by a linear regression algorithm, and calculating the average of the heat rate trends with the same number of maintenance times as a standard heat rate trend; Calculating the dispersion of the plurality of heat rate trends with the same number of maintenance times and the standard heat rate trend, dividing the trend into a plurality of levels according to the dispersion, marking the trends, and calling the heat rate graph of the corresponding category; Comparing the heat rate trends of the same steam turbine unit at the nth maintenance times and the n+1th maintenance times according to the cosine similarity algorithm, if the comparison result is less than the threshold of the heat rate trend, calling the heat rate map of the steam turbine unit corresponding to the n+1th maintenance times; The first weight optimization module of the heat rate map calls the first heat rate influencing data according to the heat rate map of the same category, extracts the first key feature and the second key feature of the first heat rate influencing data through the heat rate parameter analysis model, and optimizes the first weight of the heat rate map; The second weight optimization module of the heat rate map calls the second heat rate influencing data according to the heat rate map of the same category, extracts the third key feature and the fourth key feature of the second heat rate influencing data through the heat rate parameter analysis model, and optimizes the second weight of the heat rate map; The heat rate map evaluation module calls the corresponding first heat rate impact data and second heat rate impact data for evaluation according to the heat rate map.

2. The steam turbine unit heat rate evaluation system based on historical data according to claim 1, characterized in that: Classifying the historical heat rate data of multiple groups of steam turbine units within a T period includes: dividing the historical heat rate data based on the historical heat rate data of all the steam turbine units with maintenance time T as a dividing point, and classifying the data according to the total number of maintenance times T as the number of maintenance times.

3. The steam turbine unit heat rate evaluation system based on historical data according to claim 1, characterized in that: Calling the first heat rate influencing data according to the heat rate map of the same category, extracting the first key feature and the second key feature of the first heat rate influencing data through a heat rate parameter analysis model, and optimizing the first weight of the heat rate map includes: The first key feature and the second key feature are obtained according to an aging impact analysis model of the heat rate parameter analysis model; the aging impact analysis model includes an impact parameter preprocessing module, an impact parameter feature analysis module, a first weight optimization module and a first weight storage module; The influencing parameter preprocessing module performs data segmentation and data standardization on the first heat rate influencing data to generate pre-input data; the influencing parameter feature analysis module predicts the data of the next segmented time period through 1 layer of Transformer and 6 layers of LSTM, and identifies the first key feature of the pre-input data and the second key feature of the data of the next segmented time period through 2 layers of residual convolution layers, 1 layer of dilated convolution layers and 4 layers of LSTM; The first weight optimization module optimizes the associated weight of the heat rate map according to the first key feature and the second key feature; and the first weight storage module stores the associated weight of the heat rate map.

4. The steam turbine unit heat rate evaluation system based on historical data according to claim 3, characterized in that: The first weight optimization module optimizes the association weight of the heat rate map according to the first key feature and the second key feature, including: in, is the first weight after optimization of the i-th first heat rate influencing data, is the initial association weight of the i-th first heat rate influencing data, The optimization parameters of the first key feature obtained by the aging impact analysis model, The optimization parameters of the second key feature obtained by the aging impact analysis model; Key feature level labels.

5. The steam turbine unit heat rate evaluation system based on historical data according to claim 4, characterized in that: The calculation of the key feature level label includes: Among them, PD is the distance function, K is the k-means clustering algorithm, is the DTW algorithm, is the key feature of the input, Calculate key characteristics for the rating, is the key feature discreteness, is the cluster center, is the cluster radius.

6. The steam turbine unit heat rate evaluation system based on historical data according to claim 1, characterized in that: Calling the second heat rate influencing data according to the heat rate map of the same category, extracting the third key feature and the fourth key feature of the second heat rate influencing data through the heat rate parameter analysis model, and optimizing the second weight of the heat rate map includes: The third key feature and the fourth key feature are obtained according to the abnormality identification analysis model of the heat rate parameter analysis model; the abnormality identification analysis model includes an abnormal parameter preprocessing module, an abnormal parameter feature identification module, a second weight optimization module and a second weight storage module; The abnormal parameter preprocessing module performs data segmentation and data standardization on the second heat rate impact data to generate pre-identification data; the abnormal parameter feature recognition module extracts the third key feature of the pre-identification data based on 1 layer of Transformer, 1 layer of time convolution layer and 4 layers of LSTM units, and extracts the fourth key feature of the pre-identification data through 2 layers of residual convolution layers, 1 layer of expansion convolution layer and 4 layers of LSTM; The second weight optimization module optimizes the associated weight of the second heat rate influencing data based on the third key feature and the fourth key feature; and the second weight storage module stores the optimized associated weight.

7. The steam turbine unit heat rate evaluation system based on historical data according to claim 6, characterized in that: The second weight optimization module optimizes the associated weight of the second heat rate influencing data based on the third key feature and the fourth key feature, including: in, is the second weight after the j-th second heat rate influencing data is optimized, is the initial association weight of the jth second heat rate influencing data, is the optimization parameter of the third key feature obtained by the abnormality recognition analysis model, is the optimization parameter of the fourth key feature obtained by the abnormality recognition analysis model, and exp is an exponential function with the natural constant e as the base.

8. A method for evaluating the heat rate of a steam turbine unit based on historical data, characterized in that: include: Obtain historical heat rate data of multiple steam turbine units; Associating the first heat rate influencing data and the second heat rate influencing data according to the historical heat rate data, generating a heat rate map, and initializing an association weight of the heat rate map; The first heat rate influencing data is the influencing parameters of the steam turbine unit with aging equipment, and the second heat rate influencing data is the influencing parameters of the steam turbine unit with abnormal operation; the influencing parameters include: steam temperature, pressure and condenser vacuum; Classifying the historical heat rate data of multiple groups of steam turbine units within a T period, generating heat rate trends, comparing and marking the heat rate trends and calling the corresponding heat rate graphs; The process of comparing the heat rate trend and calling the corresponding heat rate map includes: extracting the heat rate trend of the segmented historical heat rate data by a linear regression algorithm, and calculating the average of the heat rate trends with the same number of maintenance times as a standard heat rate trend; Calculating the dispersion of the plurality of heat rate trends with the same number of maintenance times and the standard heat rate trend, dividing the trend into a plurality of levels according to the dispersion, marking the trends, and calling the heat rate graph of the corresponding category; Comparing the heat rate trends of the same steam turbine unit at the nth maintenance times and the n+1th maintenance times according to the cosine similarity algorithm, if the comparison result is less than the threshold of the heat rate trend, calling the heat rate map of the steam turbine unit corresponding to the n+1th maintenance times; calling the first heat rate influencing data according to the heat rate map of the same category, extracting the first key feature and the second key feature of the first heat rate influencing data through a heat rate parameter analysis model, and optimizing the first weight of the heat rate map; calling the second heat rate influencing data according to the heat rate map of the same category, extracting the third key feature and the fourth key feature of the second heat rate influencing data through the heat rate parameter analysis model, and optimizing the second weight of the heat rate map; During the evaluation, the first heat rate impact data and the second heat rate impact data corresponding to the heat rate map are called for evaluation.

Images