A method for analyzing heat consumption of cement kiln system based on data analysis

Abstract

The present invention discloses a method for analyzing heat consumption of a cement kiln system based on data analysis, which relates to the technical field of industrial data analysis, including collecting and preprocessing process operation parameters of a cement kiln system; based on a sliding time window and a two-stage clustering algorithm, performing time series analysis and operating condition division on the process operation parameters, and establishing an operating condition feature library; based on the operating condition feature library, establishing a heat consumption analysis base model and a heat consumption abnormal pattern recognition rule library; applying the heat consumption analysis base model to real-time operation data, and combining the heat consumption abnormal pattern recognition rule library to output a heat consumption analysis report. The present invention improves the traditional cement kiln operating condition division method by introducing a sliding time window and a two-stage clustering algorithm, which can better capture the time series correlation characteristics of the process operation parameters, dynamically adapt to system changes, and achieve more accurate operating condition division.

Description

Technical Field

[0001] The invention relates to the technical field of industrial data analysis, and in particular to a heat consumption analysis method of a cement kiln system based on data analysis. Background Art

[0002] Cement production is a major energy consumer, of which the heat consumption of the rotary kiln system accounts for more than 80% of the total energy consumption of cement production. With the deepening of industrial digital transformation, although cement companies have established a complete automated data acquisition system to collect process operation parameters such as kiln tail flue gas temperature, kiln speed, and air volume in real time through the DCS system, the existing analysis methods are too simple and cannot fully explore the deep laws contained in the data, resulting in a large amount of valuable production data not being effectively utilized.

[0003] Traditional heat consumption analysis methods are mainly based on static heat balance calculations, ignoring the characteristics of time series data, and often using a single statistical analysis or empirical model, failing to fully utilize advanced data analysis technologies such as machine learning and deep learning. At the same time, due to the lack of intelligent working condition identification methods, existing analysis often performs extensive and unified processing of data from different operating conditions, ignoring the significant impact of working condition differences on system heat consumption, which makes it difficult for cement companies to achieve accurate heat consumption prediction and optimization control, restricting the company's progress in further reducing energy consumption and improving production efficiency. Summary of the invention

[0004] In view of the shortcomings of the existing cement kiln system heat consumption analysis method in terms of working condition identification, time series characteristic analysis and heat consumption evaluation, the present invention is proposed.

[0005] Therefore, the problem to be solved by the present invention is how to achieve accurate analysis and optimal control of the heat consumption of the cement kiln system, so as to provide effective decision support for energy conservation and consumption reduction of cement enterprises.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In the first aspect, an embodiment of the present invention provides a heat consumption analysis method for a cement kiln system based on data analysis, which includes collecting and preprocessing process operation parameters of the cement kiln system, the process operation parameters including temperature parameters, flow parameters and production parameters; based on a sliding time window and a two-stage clustering algorithm, performing time series analysis and operating condition division on the process operation parameters, and establishing an operating condition feature library; based on the operating condition feature library, establishing a heat consumption analysis base model and a heat consumption abnormal pattern recognition rule library; applying the heat consumption analysis base model to real-time operation data, combining the heat consumption abnormal pattern recognition rule library, and outputting a heat consumption analysis report.

[0008] As a preferred solution of the cement kiln system heat consumption analysis method based on data analysis described in the present invention, the establishment of an operating condition feature library includes the following steps: using a sliding time window to segment the process operation parameters, setting the window length and overlap rate, and obtaining time series data segments; calculating the operating condition characteristics of temperature parameters, flow parameters and production parameters in each time series data segment; organizing the operating condition characteristics in a preset order to form a feature vector, and performing dimensionality reduction processing; using a two-stage clustering algorithm to classify the feature vector to obtain an operating condition feature cluster; calculating the heat consumption characteristic index of each operating condition feature cluster, analyzing the conversion relationship between the operating condition feature clusters, and establishing an operating condition conversion sequence; sorting and storing the operating condition feature cluster data to establish an operating condition feature library.

[0009] As a preferred solution of the cement kiln system heat consumption analysis method based on data analysis described in the present invention, the method comprises the following steps: calculating the distance matrix between the feature vectors and performing hierarchical clustering to obtain an initial cluster cluster; optimizing the initial cluster cluster by using the DBSCAN density clustering algorithm to obtain an operating condition characteristic cluster; evaluating and verifying the rationality of the obtained operating condition characteristic cluster; calculating the distance matrix between the feature vectors and performing hierarchical clustering, including calculating the weighted distance between any two feature vectors to obtain a weighted distance matrix D; treating each feature vector as an independent cluster cluster, and iteratively merging the two nearest classes by using the Ward method based on the weighted distance matrix D; determining the optimal cluster number K by using the Gap statistic; and cutting the hierarchical tree according to the optimal cluster number K to obtain K initial cluster clusters.

[0010] As a preferred solution of the cement kiln system heat consumption analysis method based on data analysis described in the present invention, wherein: using the DBSCAN density clustering algorithm to optimize the initial clustering clusters includes the following steps: calculating the local density of the sample points in each initial clustering cluster, and selecting the local density peak point as the clustering center; according to the density accessibility principle, each sample point is assigned to the nearest clustering center; after the sample point assignment is completed, the density similarity of all clustering clusters is checked, and if two clustering clusters meet the density similarity standard and meet the working condition conversion smoothness constraint, the two clustering clusters are merged; for unassigned sample points, the nearest neighbor principle is used to assign them to the nearest clustering cluster; the calculation formula of the local density is as follows:

[0011] ,

[0012] in, is the local density of sample point o, o is the sample point to be calculated, is the mth sample point in the neighborhood, For sample points The energy gradient at is the energy consumption gradient weight coefficient, is the Gaussian kernel function.

[0013] As a preferred solution of the cement kiln system heat consumption analysis method based on data analysis described in the present invention, the establishment of the operating condition conversion sequence includes the following steps: based on the time series data of the operating condition feature cluster, calculating the operating condition conversion matrix T, wherein the operating condition conversion matrix element T ab Represents the number of conversions from operating condition feature cluster a to operating condition feature cluster b; calculate each non-zero operating condition conversion matrix element T ab The corresponding conversion success rate; construct the working condition conversion network, with each working condition feature cluster as a node and the working condition conversion matrix element T ab is the edge weight; based on the operating condition conversion network, the high-frequency conversion paths are extracted, the typical sequence patterns of the operating condition conversion are analyzed, and the operating condition conversion sequence is formed.

[0014] As a preferred solution of the heat consumption analysis method of the cement kiln system based on data analysis described in the present invention, the establishment of a heat consumption analysis base model and a heat consumption abnormal pattern recognition rule base includes the following steps: exporting historical data of the operating condition feature cluster from the operating condition feature library to establish a heat consumption analysis data set; performing time series division on the heat consumption analysis data set to obtain an analysis data matrix and a verification data matrix; inputting the analysis data matrix into a multi-channel analysis network to obtain a heat consumption analysis base model; establishing a heat consumption abnormal pattern recognition rule base based on the verification data matrix, and constructing a heat consumption analysis result interpretation mechanism.

[0015] As a preferred solution of the heat consumption analysis method of the cement kiln system based on data analysis described in the present invention, the analysis data matrix is ​​input into the multi-channel analysis network to obtain the heat consumption analysis base model, which includes the following steps: constructing an operating condition feature channel according to the analysis data matrix to obtain the operating condition time series feature representation; constructing an operating condition conversion channel according to the operating condition conversion sequence and the operating condition time series feature representation in the analysis data matrix to obtain the operating condition conversion impact feature; constructing a time feature channel according to the time series information in the analysis data matrix to obtain the heat consumption cycle feature; constructing a feature adaptive fusion layer according to the operating condition time series feature representation, the operating condition conversion impact feature and the heat consumption cycle feature to obtain a comprehensive feature representation; constructing a heat consumption change feature extraction module according to the comprehensive feature representation to obtain the heat consumption analysis base model.

[0016] In the second aspect, an embodiment of the present invention provides a heat consumption analysis system for a cement kiln system based on data analysis, which includes a data processing module for collecting and preprocessing process operation parameters of the cement kiln system; a feature library construction module for performing time series analysis and operating condition division on the process operation parameters based on a sliding time window and a two-stage clustering algorithm, and establishing an operating condition feature library; a model construction module for establishing a heat consumption analysis base model and a heat consumption abnormal pattern recognition rule library based on the operating condition feature library; a report generation module for applying the heat consumption analysis base model to real-time operation data, and outputting a heat consumption analysis report in combination with the heat consumption abnormal pattern recognition rule library.

[0017] The beneficial effects of the present invention are as follows: the present invention improves the traditional cement kiln working condition division method by introducing a sliding time window and a two-stage clustering algorithm, which can better capture the time series correlation characteristics of process operation parameters, dynamically adapt to system changes, and achieve more accurate working condition division. At the same time, a heat consumption analysis base model is established based on the working condition feature library, which comprehensively considers the working condition time series characteristics, working condition conversion impact characteristics and heat consumption cycle characteristics, and improves the analysis ability of heat consumption changes. In addition, the proposed heat consumption abnormal pattern recognition rule library and interpretation mechanism can effectively identify heat consumption abnormalities and provide improvement suggestions to help optimize the process and energy-saving control. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. 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 creative work.

[0019] Figure 1 The framework flow chart of the heat consumption analysis method of cement kiln system based on data analysis.

[0020] Figure 2 A flow chart is constructed for the operating condition feature library of the heat consumption analysis method of cement kiln system based on data analysis.

[0021] Figure 3 Flow chart of the initial clustering optimization process for the heat consumption analysis method of cement kiln system based on data analysis. DETAILED DESCRIPTION

[0022] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0023] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0024] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0025] Example 1, reference Figure 1~Figure 3 , which is the first embodiment of the present invention, and provides a cement kiln system heat consumption analysis method based on data analysis, the framework steps are as follows Figure 1 As shown, including

[0026] S1: Collect and pre-process the process operation parameters of the cement kiln system.

[0027] Specifically, the process operation parameters include temperature parameters, flow parameters and production parameters, among which temperature parameters include the flue gas temperature at the kiln tail, the temperature of the decomposition zone in the kiln and the temperature at the kiln head, flow parameters include kiln speed, primary air volume and secondary air volume, and production parameters include the amount of raw materials entering the kiln, clinker output and coal powder consumption. The selection of these process operation parameters is based on the characteristics of the thermal process of the cement kiln system, among which temperature parameters reflect the heat distribution characteristics of the system, flow parameters characterize the transfer law of materials and heat, and production parameters reflect the energy balance state of the system.

[0028] Furthermore, the sampling frequency of process operation parameters is determined according to the dynamic characteristics of each parameter: fast-changing parameters (such as kiln head temperature and primary air volume) use a sampling frequency of 10 seconds / time, other temperature parameters and flow parameters use a sampling frequency of 30 seconds / time, and production parameters use a sampling frequency of 1 minute / time. The data quality assessment of process operation parameters adopts the following standards: data availability is not less than 95%, and the data fluctuation range should be between 30% and 70% of the equipment range.

[0029] Furthermore, considering that the process operation parameters in the actual production process may have abnormal values ​​and missing values, the present invention first pre-processes the collected process operation parameters, specifically including: using 3 The missing values ​​were filled by linear interpolation, the data were smoothed by sliding average method, and finally the Z-score was normalized.

[0030] S2: Based on the sliding time window and two-stage clustering algorithm, the process operation parameters are analyzed in time series and the operating conditions are divided to establish an operating condition feature library.

[0031] Currently, the working condition division of cement kiln system mostly adopts fixed threshold judgment or simple clustering algorithm. These methods do not consider the time series correlation characteristics of process operation parameters, and the working condition division boundary is fixed, which is difficult to adapt to the dynamic changes of the system. In view of the above problems, the present invention proposes a working condition division method based on sliding time window and two-stage clustering. The working condition feature library establishment flow chart is as follows: Figure 2 As shown, the specific steps include:

[0032] S2.1: Use a sliding time window to segment the process operation parameters, set the window length and overlap rate, and obtain time series data segments.

[0033] Preferably, the window length is set adaptively: the minimum window length is determined based on the autocorrelation analysis of the process operation parameters, the maximum window length is determined by Fourier transform analysis of the energy consumption fluctuation period, and the grid search method is used to determine the optimal window length within this interval. The overlap rate is fixedly set to 50%, and the number of valid data points in each time window should be no less than 90% of the total data points in the window.

[0034] S2.2: Calculate the operating characteristics of temperature parameters, flow parameters and production parameters in each time series data segment.

[0035] Specifically, the operating condition characteristics include statistical characteristics (such as mean, standard deviation), dynamic characteristics (such as change rate, fluctuation frequency) and time series characteristics (such as trend characteristics, autocorrelation coefficient). In addition, entropy-based operating condition complexity characteristics are introduced, including sample entropy and approximate entropy, to characterize the stability of the operating condition.

[0036] S2.3: Organize the working condition characteristics in a preset order to form a feature vector, and perform dimensionality reduction processing.

[0037] Specifically, the maximum information coefficient (MIC) was first used to evaluate the nonlinear correlation of features on heat consumption, and features with MIC values ​​greater than 0.3 were selected; then principal component analysis (PCA) was used for preliminary dimensionality reduction, and the principal components with a cumulative contribution rate of 85% were retained; finally, Lasso regression was used for feature screening, and features with non-zero regression coefficients were selected as the final feature vectors.

[0038] S2.4: Use a two-stage clustering algorithm to classify the feature vectors and obtain the operating condition feature clusters.

[0039] Specifically, the following steps are included:

[0040] S2.4.1: Calculate the distance matrix between feature vectors and perform hierarchical clustering to obtain initial clusters.

[0041] Specifically, the weighted distance between any two eigenvectors is calculated to obtain a weighted distance matrix D, where the weighted distance formula is:

[0042] ,

[0043] in, is the weighted distance between the i-th eigenvector and the j-th eigenvector, are the i-th eigenvector and the j-th eigenvector respectively, are the energy consumption change rates at the corresponding moments of the i-th eigenvector and the j-th eigenvector, is the energy consumption sensitivity coefficient, and its initial value is set to 0.3.

[0044] Furthermore, each feature vector is regarded as an independent cluster; based on the weighted distance matrix D, the Ward method is used to iteratively merge the two closest classes, specifically including: calculating the intra-class error square sum SSE increment after merging any two classes; selecting the two classes with the smallest SSE increment for merging; updating the distance matrix D, and calculating the distance between the newly generated class and other classes.

[0045] Furthermore, the Gap statistic is used to determine the optimal number of clusters K, which specifically includes: calculating the intra-class dispersion W of the actual data for different cluster numbers k k ; Generate B random reference data sets and calculate the intra-class dispersion W of each reference data set kb ; Calculate the Gap statistic and standard deviation s k , the specific formula is as follows:

[0046] ,

[0047] ,

[0048] Where B is the number of random reference datasets, W k is the intra-class dispersion of the actual data, W kn is the intra-class dispersion of the nth random reference data set, is the standard deviation, for The mean of , n is the index, and k is the number of clusters.

[0049] Further, choose to meet The minimum k value is taken as the optimal clustering number K; the hierarchical tree is cut according to the determined optimal clustering number K to obtain K initial clusters.

[0050] S2.4.2: Use the improved DBSCAN density clustering algorithm to optimize the initial clustering clusters and obtain the working condition feature clusters.

[0051] Specifically, the initial clustering optimization process flow chart is as follows Figure 3 As shown in the figure, the local density of each sample point in the initial cluster is calculated; after the local density calculation is completed, the local density peak point is selected as the cluster center; for each sample point, according to the density accessibility principle, it is assigned to the nearest cluster center; in the process of sample point allocation, the energy consumption change value of two adjacent clusters is calculated. If it exceeds the preset energy consumption change threshold, the operating condition transition smoothness constraint is violated and the two clusters are prohibited from being merged; after the sample point allocation is completed, the density similarity of all clusters is checked. If two clusters meet the density similarity standard and the operating condition transition smoothness constraint, the two clusters are merged; for unassigned sample points, the nearest neighbor principle is used to assign them to the nearest cluster. After the above steps, the optimization of the initial cluster is completed, and the optimized operating condition feature cluster is obtained.

[0052] Preferably, the calculation formula of local density is as follows:

[0053] ,

[0054] in, is the local density of sample point o, o is the sample point to be calculated, is the mth sample point in the neighborhood, For sample points The energy gradient at is the energy consumption gradient weight coefficient, is the Gaussian kernel function, defined as:

[0055] ,

[0056] in, is the bandwidth parameter of the kernel function.

[0057] It should be noted that the absolute value of the energy consumption gradient is introduced into the calculation formula of the local density in order to increase the density of the area with high energy consumption changes, which is conducive to identifying the working condition characteristics of energy consumption mutations. The density accessibility principle means that one sample point can be reached from another sample point through several sample points with similar density. The preset energy consumption change threshold is set to 3 times the standard deviation of the energy consumption change rate at adjacent moments. The criterion for determining density similarity is that the average local density difference between the two clusters is less than the density similarity threshold, and the density similarity threshold is set to 2 times the standard deviation of the local density of the sample point.

[0058] S2.4.3: Evaluate and verify the rationality of the obtained operating condition characteristic clusters.

[0059] Specifically, the Davies-Bouldin index (DBI) and silhouette coefficient (SC) are used to evaluate the clustering effect, where the calculation formula of the Davies-Bouldin index is:

[0060] ,

[0061] Where N is the number of operating condition feature clusters, are the average radii of the a-th operating condition feature cluster and the b-th operating condition feature cluster, respectively. is the distance between the centers of the working condition feature clusters, and a and b are both indexes.

[0062] Furthermore, the evaluation criteria are DBI < 0.6 and SC > 0.7. If the evaluation criteria are not met, the neighborhood radius parameter and MinPts parameter of the density clustering algorithm are adjusted for optimization, and the parameter size is adjusted according to the direction of the indicator deviation until the evaluation criteria are met or the maximum number of iterations (set to 30 times) is reached.

[0063] Furthermore, rationality verification includes inter-class significance verification, process feasibility verification and heat consumption impact verification. Among them, inter-class significance verification is performed by calculating the Mahalanobis distance between the operating condition feature clusters for significance test, and a p value less than 0.05 indicates significant differences between classes; process feasibility verification evaluates the feasibility of process parameter combinations according to the fuzzy rules of the expert knowledge base to ensure that the parameter combinations of each operating condition feature cluster meet the production constraints; heat consumption impact verification evaluates the rationality of operating condition conversion by analyzing the impact of operating condition conversion on heat consumption indicators, and adjusts or merges the operating condition feature clusters that do not meet the requirements.

[0064] S2.5: Based on the obtained N operating condition characteristic clusters, calculate the heat consumption characteristic index of each operating condition characteristic cluster.

[0065] Specifically, the heat consumption characteristic indicators include the heat consumption per unit product of standard coal, heat utilization rate, system energy efficiency coefficient and dynamic heat consumption change rate.

[0066] S2.6: Analyze the conversion relationship between operating condition feature clusters and establish an operating condition conversion sequence.

[0067] Specifically, based on the time series data of the operating condition feature cluster, the operating condition conversion matrix T is calculated, where the operating condition conversion matrix element T ab represents the number of conversions from operating condition feature cluster a to operating condition feature cluster b; for non-zero operating condition conversion matrix elements T ab , calculate the corresponding conversion success rate; construct the working condition conversion network, take each working condition feature cluster as node, and take the working condition conversion matrix element T abThe edge weights are taken as the edge weights, and invalid conversions with conversion success rates lower than the preset confidence level are eliminated to obtain the optimized operating condition conversion network. Based on the optimized operating condition conversion network, the depth-first search algorithm is used to extract the paths with conversion frequencies in the top I% as high-frequency conversion paths, where I is the preset percentage threshold. For each high-frequency conversion path, its total conversion success rate is verified to meet the process requirements, and its feasibility is verified based on energy consumption fluctuations, equipment parameter constraints, etc., to finally form an operating condition conversion sequence.

[0068] S2.7: Use a hierarchical structure to organize and store the operating condition feature cluster data and establish an operating condition feature library.

[0069] Specifically, a three-layer storage structure is adopted: the first layer stores the basic information of N operating condition feature clusters, including the number of the operating condition feature cluster and the representative process operation parameter values ​​in each operating condition feature cluster; the second layer stores the heat consumption characteristic index; the third layer stores the operating condition conversion sequence. The above-mentioned hierarchical data are organized and stored in the database in a unified format to form a complete operating condition feature library for subsequent operating condition identification and optimization. The operating condition feature library data is updated once every production cycle (usually 1 month), and the data of the last three cycles are retained for analyzing the trend of operating condition changes.

[0070] S3: Based on the operating condition feature library, a basic model for heat consumption analysis is established.

[0071] Specifically, the following steps are included:

[0072] S3.1: Export the historical data of the operating condition feature cluster from the operating condition feature library and establish a heat consumption analysis data set.

[0073] Specifically, the process operation parameter values, heat consumption characteristic indicators, and operating condition conversion sequences of each operating condition characteristic cluster are extracted from the operating condition characteristic library to obtain a heat consumption analysis data set.

[0074] S3.2: Perform time series division on the heat consumption analysis data set to obtain the analysis data matrix and the verification data matrix.

[0075] Specifically, the time series cross-validation method is used to divide the modeling data set, set the sliding window length and analysis step size; for each sliding window, the first 70% of the data constitutes the analysis data matrix, and the last 30% of the data constitutes the verification data matrix; in view of the uneven distribution of samples of different working condition feature clusters, the data resampling method is used to ensure that the proportion of each working condition feature cluster in the analysis data matrix and the verification data matrix is ​​consistent. In order to maintain the continuity of the time series features, resampling is only performed between adjacent time points.

[0076] S3.3: Input the analysis data matrix into the multi-channel analysis network to construct the heat consumption analysis base model.

[0077] Specifically, the following steps are included:

[0078] S3.3.1: Construct the operating condition characteristic channel based on the analysis data matrix and obtain the operating condition time series characteristic representation.

[0079] Specifically, a double-layer structure of a short-term feature extraction layer and a long-term feature extraction layer is used to extract time series features; the short-term feature extraction layer sets the sliding window length to 4 hours and the step length to 10 minutes, and extracts the mean, standard deviation, change rate and other statistics of the operating parameters within the window; the long-term feature extraction layer sets a feature extraction window of 24 hours in length to calculate the statistical features of the daily mean, daily peak, daily fluctuation range and other statistical features of the operating parameters; the statistics obtained by the short-term feature extraction layer are concatenated and fused with the statistical features obtained by the long-term feature extraction layer, and the working condition time series feature representation is output.

[0080] S3.3.2: Based on the operating condition conversion sequence and operating condition timing characteristics in the analysis data matrix, construct the operating condition conversion channel and obtain the operating condition conversion impact characteristics.

[0081] Specifically, firstly, the operating condition conversion characteristics are extracted from the analysis data matrix. The operating condition conversion characteristics include the operating condition conversion time interval, the operating condition conversion frequency, and the parameter change amplitude of the operating condition conversion. Among them, the operating condition conversion time interval is obtained by calculating the time difference between two adjacent operating condition conversions, the operating condition conversion frequency is obtained by the number of operating condition conversions per unit time, and the parameter change amplitude of the operating condition conversion is obtained by the parameter difference before and after the operating condition conversion. Then, an impact attenuation model of the operating condition conversion is constructed. The impact attenuation model describes the relationship between the degree of influence of the operating condition conversion on the heat consumption and time. The model can characterize the heat consumption fluctuation characteristics caused by the operating condition conversion. Finally, the operating condition conversion characteristics are combined with the characteristic parameters of the impact attenuation model to form the operating condition conversion impact characteristics.

[0082] S3.3.3: Based on the time series information in the analysis data matrix, construct a time characteristic channel to obtain the heat consumption cycle characteristics.

[0083] Specifically, wavelet decomposition is first used to decompose the time series into periodic components of different scales, such as 24 hours, 7 days, and 30 days. Then, Fourier transform is performed on the components of each scale to obtain their amplitude spectrum and phase spectrum characteristics. Finally, the spectral features of each scale are combined to output the heat consumption periodic characteristics.

[0084] S3.3.4: Based on the operating condition timing feature representation, operating condition conversion impact features and heat consumption cycle features, a feature adaptive fusion layer is constructed to obtain a comprehensive feature representation.

[0085] Specifically, the mutual information method is first used to calculate the correlation scores between the three types of features and the heat consumption data to obtain the feature importance evaluation results; then the relevant weight matrix between the three types of features is calculated based on the attention mechanism; finally, the feature importance evaluation results are combined with the relevant weight matrix to generate fusion weights, and the fusion weights are used to perform weighted combination of the three types of features to output a comprehensive feature representation.

[0086] S3.3.5: Based on the comprehensive feature representation, construct a heat consumption change feature extraction module to obtain the heat consumption analysis base model.

[0087] Specifically, polynomial fitting is first used to extract the trend characteristics of the heat consumption series. The order of the polynomial is selected according to the AIC criterion, and the slope and curvature of the trend line are calculated to identify the trend inflection point. Then, wavelet transform is used to detect the mutation characteristics of the heat consumption series, and the fluctuation characteristics such as fluctuation period and fluctuation amplitude are calculated. Finally, based on the mutual information coefficient and Granger causality test method, the correlation characteristics between the heat consumption and the operating parameters are extracted. The trend characteristics, fluctuation characteristics and correlation characteristics are combined to form a heat consumption change feature extraction module, thereby constructing the heat consumption analysis base model.

[0088] S3.4: Based on the verification data matrix, establish a heat consumption anomaly pattern recognition rule base.

[0089] Specifically, the verification data matrix is ​​analyzed using the heat consumption analysis base model to obtain the heat consumption change characteristics; based on expert knowledge and statistical analysis methods, the heat consumption change characteristics are correlated with process parameters, equipment status, etc., and heat consumption abnormal pattern recognition rules are established; the rules include but are not limited to: heat consumption mutation rules, heat consumption slow change rules, periodic fluctuation rules, etc. The rules contain specific judgment thresholds and confidence indicators.

[0090] S3.5: Construct a mechanism for interpreting heat consumption analysis results.

[0091] Specifically, based on the heat rate abnormal pattern recognition rule base, a result interpretation template is designed, including: description of heat rate change trend, abnormal pattern determination results, correlation factor analysis, improvement suggestions, etc. For different heat rate abnormal patterns, combined with historical cases in the operating condition feature library, corresponding interpretation strategies and improvement suggestion generation methods are established.

[0092] S4: Apply the heat consumption analysis base model to the real-time operation data, combine the heat consumption anomaly pattern recognition rule base and result interpretation mechanism, and output the heat consumption analysis report.

[0093] Specifically, the heat consumption analysis report includes: heat consumption change characteristic analysis, abnormal pattern recognition results, correlation factor analysis results, specific improvement suggestions, etc. The analysis report contains both quantitative analysis results and qualitative analysis conclusions, providing decision support for process optimization and energy-saving improvements.

[0094] Furthermore, the present embodiment also provides a heat consumption analysis system for a cement kiln system based on data analysis, including a data processing module for collecting and preprocessing process operation parameters of the cement kiln system; a feature library construction module for performing time series analysis and operating condition division on the process operation parameters based on a sliding time window and a two-stage clustering algorithm, and establishing an operating condition feature library; a model construction module for establishing a heat consumption analysis base model and a heat consumption abnormal pattern recognition rule library based on the operating condition feature library; and a report generation module for applying the heat consumption analysis base model to real-time operation data, and outputting a heat consumption analysis report in combination with the heat consumption abnormal pattern recognition rule library.

[0095] In summary, the present invention improves the traditional cement kiln working condition division method by introducing a sliding time window and a two-stage clustering algorithm, which can better capture the time series correlation characteristics of process operation parameters, dynamically adapt to system changes, and achieve more accurate working condition division. At the same time, a heat consumption analysis base model is established based on the working condition feature library, which comprehensively considers the working condition time series characteristics, working condition conversion impact characteristics and heat consumption cycle characteristics, and improves the analysis ability of heat consumption changes. In addition, the proposed heat consumption abnormal pattern recognition rule library and interpretation mechanism can effectively identify heat consumption anomalies and provide improvement suggestions to help optimize process and energy-saving control.

[0096] Example 2, reference Figure 1~Figure 3 , which is the second embodiment of the present invention, and this embodiment provides a cement kiln system heat consumption analysis method based on data analysis. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0097] Temperature parameters such as the flue gas temperature at the kiln tail, the temperature of the decomposition zone in the kiln, and the temperature at the kiln head play a key role in simulating the heat distribution characteristics in the kiln. Taking the actual production data of a cement enterprise as an example, the average flue gas temperature at the kiln tail is 285℃, the average temperature of the decomposition zone in the kiln is 1100℃, and the average temperature at the kiln head is 175℃. The kiln speed is about 3r / min, and the average primary air volume is 28000Nm 3 / h, the average secondary air volume is 37000Nm 3 / h. The average raw material volume entering the kiln is 200t / h, the average clinker output is 160t / h, and the average coal powder consumption is 18.5t / h. Based on the above process operation parameter data, it is pre-processed for subsequent time series analysis and working condition division.

[0098] The data were analyzed by sliding time window and two-stage clustering algorithm, and a total of 6 typical operating condition characteristic clusters were identified, and some data are shown in Table 1. The characteristic values ​​of the 6 operating condition characteristic clusters in temperature parameters, flow parameters and production parameters are shown in Table 1. Among them, operating condition 1 represents a stable production state with low kiln temperature and small flow parameters; operating condition 2 represents a normal production state with medium kiln temperature and large flow; operating condition 3 represents a high-load production state with high kiln temperature and large flow parameters; operating condition 4 represents an unstable state with large kiln temperature fluctuations and frequent changes in flow parameters; operating condition 5 represents an energy-saving production state with low kiln temperature and small coal powder consumption; operating condition 6 represents a high-energy-consuming production state with high kiln temperature and large coal powder consumption. The data distribution of the 6 operating condition characteristic clusters is relatively balanced, and the clustering effect is good as verified by DBI and SC indicators.

[0099] Table 1 Characteristic clusters of some typical working conditions

[0100] Condition No. Kiln tail flue gas temperature Decomposition zone temperature in kiln Kiln head temperature Kiln speed ... Pulverized coal consumption Condition 1 260℃ 1050℃ 180℃ 2.8r / min ... 18t / h Condition 2 300℃ 1100℃ 200℃ 3.0r / min ... 19t / h Condition 3 340℃ 1150℃ 220℃ 3.2r / min ... 22t / h Condition 4 280~320℃ 1080-1120℃ 190~210℃ 2.8~3.2r / min ... 18~22t / h Condition 5 280℃ 1070℃ 190℃ 2.9r / min ... 16t / h Condition 6 320℃ 1130℃ 210℃ 3.1r / min ... 24t / h

[0101] As can be seen from Table 1, the present invention captures the different operating characteristics of the cement kiln system through comprehensive analysis of multiple process parameters, providing basic support for subsequent heat consumption analysis. Compared with the existing operating condition division method that only relies on a single parameter or simple clustering, the method of the present invention using a sliding time window and a two-stage clustering algorithm can better reflect the dynamic change characteristics of the system and quantify the transition properties between operating conditions, so it can more accurately identify and characterize the heat consumption characteristics of the cement kiln system.

[0102] Based on the above-mentioned operating condition feature library, the present invention constructs a heat consumption analysis base model, which is verified on the actual production data from January to June 2022. The results show that the model can better capture the trend change characteristics of the kiln system's heat consumption, the periodic fluctuation characteristics, and the impact of operating condition conversion on heat consumption. For example, an abnormal fluctuation in heat consumption occurred from late March to early April 2022. The model accurately identified that the main cause was the drastic changes in kiln speed, primary air volume, and coal powder consumption, which led to drastic fluctuations in the temperature field in the kiln, and then caused abnormal fluctuations in heat consumption. The model gives targeted improvement suggestions through correlation analysis, such as adjusting the air volume control strategy, optimizing the coal powder supply system, etc. These suggestions have received positive responses on site and achieved good results. Overall, the heat consumption analysis method of the present invention is significantly superior to existing empirical analysis methods in terms of accuracy, applicability, and interpretability.

[0103] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A cement kiln system heat consumption analysis method based on data analysis, characterized in that: include, Collecting and preprocessing process operation parameters of the cement kiln system, wherein the process operation parameters include temperature parameters, flow parameters and production parameters; Based on a sliding time window and a two-stage clustering algorithm, the process operation parameters are analyzed in time series and divided into working conditions, and a working condition feature library is established; Based on the operating condition feature library, a heat consumption analysis base model and a heat consumption abnormal pattern recognition rule library are established; Apply the heat consumption analysis base model to real-time operation data, combine it with the heat consumption abnormal pattern recognition rule base, and output the heat consumption analysis report; The establishment of the operating condition feature library comprises the following steps: Use sliding time windows to segment process operation parameters, set window length and overlap rate, and obtain time series data segments; Calculate the operating characteristics of temperature parameters, flow parameters and production parameters in each time series data segment; Organizing the working condition characteristics in a preset order to form a feature vector, and performing dimensionality reduction processing; A two-stage clustering algorithm is used to classify the feature vectors to obtain operating condition feature clusters; Calculate the heat consumption characteristic index of each operating condition characteristic cluster, analyze the conversion relationship between operating condition characteristic clusters, and establish the operating condition conversion sequence; Arrange and store the working condition characteristic cluster data to establish a working condition characteristic library; The method of using a two-stage clustering algorithm to classify the feature vectors comprises the following steps: Calculate the distance matrix between feature vectors and perform hierarchical clustering to obtain initial clusters; The DBSCAN density clustering algorithm is used to optimize the initial clustering cluster to obtain the working condition characteristic cluster; Evaluate and verify the rationality of the obtained operating condition characteristic cluster; The calculation of the distance matrix between the feature vectors and the hierarchical clustering includes: Calculate the weighted distance between any two eigenvectors to obtain the weighted distance matrix D; Each feature vector is regarded as an independent cluster, and the Ward method is used to iteratively merge the two closest clusters based on the weighted distance matrix D. Use the Gap statistic to determine the optimal number of clusters K; Cutting the hierarchical tree according to the optimal clustering number K to obtain K initial clusters; The optimization of the initial cluster clusters by using the DBSCAN density clustering algorithm comprises the following steps: Calculate the local density of each sample point in the initial cluster, and select the local density peak point as the cluster center; According to the density accessibility principle, each sample point is assigned to the nearest cluster center; After the sample points are assigned, the density similarity of all clusters is checked. If two clusters meet the density similarity criteria and the working condition transition smoothness constraint, the two clusters are merged. For unassigned sample points, the nearest neighbor principle is used to assign them to the nearest cluster; The calculation formula of the local density is as follows: in, is the local density of sample point o, o is the sample point to be calculated, is the mth sample point in the neighborhood, For sample points The energy gradient at is the energy consumption gradient weight coefficient, is the Gaussian kernel function, and m is the sample point index.

2. The heat consumption analysis method of cement kiln system based on data analysis according to claim 1, characterized in that: The establishment of the operating mode conversion sequence comprises the following steps: Based on the time series data of the operating condition feature cluster, the operating condition conversion matrix T is calculated, where the operating condition conversion matrix element T ab Indicates the number of conversions from operating condition feature cluster a to operating condition feature cluster b; Calculate each non-zero operating condition transformation matrix element T ab The corresponding conversion efficiency; Construct a working condition conversion network, with each working condition feature cluster as a node and the working condition conversion matrix element T ab is the edge weight; Based on the operating condition conversion network, high-frequency conversion paths are extracted, typical sequence patterns of operating condition conversions are analyzed, and an operating condition conversion sequence is formed.

3. The heat consumption analysis method of cement kiln system based on data analysis according to claim 1, characterized in that: The establishment of the heat consumption analysis base model and the heat consumption abnormal pattern recognition rule base comprises the following steps: Exporting historical data of the operating condition feature cluster from the operating condition feature library to establish a heat consumption analysis data set; Performing time series division on the heat consumption analysis data set to obtain an analysis data matrix and a verification data matrix; Inputting the analysis data matrix into a multi-channel analysis network to obtain a heat consumption analysis base model; Based on the verification data matrix, a heat consumption anomaly pattern recognition rule base is established, and a heat consumption analysis result interpretation mechanism is constructed.

4. The heat consumption analysis method of a cement kiln system based on data analysis according to claim 3, characterized in that: Inputting the analysis data matrix into a multi-channel analysis network to obtain a heat consumption analysis basic model comprises the following steps: Construct the working condition characteristic channel according to the analysis data matrix to obtain the working condition time series characteristic representation; According to the working condition conversion sequence in the analysis data matrix and the working condition time series characteristic representation, a working condition conversion channel is constructed to obtain the working condition conversion influence characteristics; According to the time series information in the analysis data matrix, the time characteristic channel is constructed to obtain the heat consumption cycle characteristics; According to the operating condition time series feature representation, the operating condition conversion impact feature and the heat consumption cycle feature, a feature adaptive fusion layer is constructed to obtain a comprehensive feature representation; According to the comprehensive feature representation, a heat consumption change feature extraction module is constructed to obtain a heat consumption analysis base model.

5. A cement kiln system heat consumption analysis system based on data analysis, based on the cement kiln system heat consumption analysis method based on data analysis according to any one of claims 1 to 4, characterized in that: Also includes, Data processing module, used to collect and pre-process the process operation parameters of the cement kiln system; A feature library building module is used to perform time series analysis and operating condition division on the process operation parameters based on a sliding time window and a two-stage clustering algorithm to establish an operating condition feature library; A model building module, used to establish a heat consumption analysis base model and a heat consumption abnormal pattern recognition rule library based on the operating condition feature library; The report generation module is used to apply the heat consumption analysis base model to the real-time operation data, combine it with the heat consumption abnormal pattern recognition rule base, and output the heat consumption analysis report.

Images