Generator set time domain correlation analysis method based on online monitoring and related equipment

Abstract

The application discloses a kind of generator set time domain correlation analysis method and related equipment based on online monitoring.The method comprises: according to the unified sampling time of generator set DCS clock, boiler operating parameter and pressure-bearing component online monitoring data are synchronously collected, and standardization time series dataset is generated after preprocessing;Standardization time series dataset is sequentially subjected to correlation preselection, time series causality test and dynamic time warping similarity matching, and effective candidate parameters with correlation and causality characteristics are filtered out layer by layer, and the action lag length of each effective candidate parameter is determined;Lasso regression model is used to extract features and calculate influence weight of effective candidate parameters, and the action lag length is combined to construct operating parameter-pressure-bearing component health state quantitative correlation matrix;According to the influence weight and action lag length of each core influence parameter in the matrix, output generator set deep peak shaving optimization control scheme.The application effectively distinguishes the accidental correlation and real physical causality of data, objectively quantifies the contribution and lag characteristics of operating parameter to pressure-bearing component thermal damage, and then realizes advanced fine regulation and control, reduces the alternating thermal stress of heated surface.

Description

Technical Field

[0001] This invention relates to the field of generator set operation monitoring and data analysis, specifically to a method and related equipment for time-domain correlation analysis of generator sets based on online monitoring. Background Technology

[0002] With the continuous expansion of wind power and photovoltaic new energy installed capacity, and the normalization of deep peak-shaving operation of domestic coal-fired power generating units, the units frequently change operating conditions within a wide load range. Operating parameters such as boiler load, furnace air volume, flue gas oxygen content, and main steam temperature and pressure fluctuate significantly and frequently. Under this operating environment, boiler pressure-bearing heat-receiving components such as water-cooled walls, superheaters, and reheaters are constantly in an alternating temperature field, repeatedly generating alternating thermal stress. Over time, this can easily lead to tube wall deformation and crack initiation, and in severe cases, bulging of the heating surface, tube rupture, and leakage, directly restricting the safe and economical operation of the unit.

[0003] Existing online monitoring systems for thermal power units typically rely on distributed control systems (DCS) and various temperature and stress sensors to independently collect boiler operating parameters and monitoring data on the temperature, expansion, and stress deformation of pressure components. Referring to the published patent CN115031256A, "Real-time Monitoring Method for Hydrodynamic Safety and Coking of Water-Cooled Walls in Once-Through Boilers," this technology relies solely on fixed-point temperature measuring elements to collect local wall temperature data, and on set thresholds and operational experience to determine the safety of the heating surface, thus only achieving single-point parameter over-limit alarms. In current boiler operation data analysis practices, most methods use fixed empirical formulas or univariate linear regression models to determine correlations, only capable of capturing simple linear variation patterns.

[0004] However, the existing technologies mentioned above are insufficient to meet the requirements of refined management in practical applications under deep peak-shaving conditions. Current boiler operating parameters and pressure-bearing component condition monitoring data belong to different acquisition links, and the two sets of monitoring data are isolated from each other. They lack a strict unified benchmark in the time dimension, resulting in a lack of systematic time-domain correlation analysis methods and an inability to establish a temporal coupling relationship between all dimensions of boiler operating parameters and the stress and fatigue damage of heated components. Furthermore, existing data analysis methods are relatively simple, lacking a hierarchical and progressive time-domain correlation analysis architecture, and cannot effectively distinguish between accidental data correlations and genuine causal disturbances. Moreover, parameter fluctuations caused by boiler operation are transmitted to pressure-bearing heated components and cause state changes, resulting in inherent time differences at the physical level. Existing analysis methods are ill-suited to the nonlinear changes, time lags, and non-stationary fluctuations of parameters under deep peak-shaving conditions. In addition, when investigating the source parameters that induce abnormal component damage, existing technologies rely entirely on staff experience to select key influencing parameters, failing to objectively extract core influencing parameters and quantify the contribution of each operating parameter to the thermal fatigue damage of pressure-bearing components. Because it is impossible to accurately locate the source parameters that induce abnormal damage to components and obtain quantified impact weights and lag durations, existing solutions are unable to guide refined control work such as load increase / decrease rates, air distribution ratios, and steam temperature optimization from a data perspective. This ultimately leads to a continuous increase in fatigue losses of the heating surface, a higher frequency of unplanned unit maintenance, and increased operation and maintenance costs. Summary of the Invention

[0005] To address the problems existing in the prior art, this invention provides a time-domain correlation analysis method and related equipment for generator sets based on online monitoring. Its purpose is to construct a hierarchical and progressive time-domain causal and lag quantitative analysis architecture, break down data silos, and objectively extract the influence weight and lag time of operating parameters on the thermal damage of pressure-bearing components.

[0006] To solve the above-mentioned technical problems, the present invention is achieved through the following technical solution: According to a first aspect of the present invention, a method for time-domain correlation analysis of generator sets based on online monitoring is provided, comprising: Based on the unified sampling time of the generator set's distributed control system (DCS) clock, the timing data of boiler operating parameters and online monitoring data of boiler pressure-bearing components are collected synchronously. The collected time-series data of boiler operating parameters and the online monitoring data of boiler pressure-bearing components are preprocessed by time alignment, anomaly removal, missing value filling and noise reduction normalization to generate a standardized time-series dataset. The standardized time-series dataset is subjected to correlation pre-selection, time-series causality test and dynamic time warping similarity matching in sequence. Valid candidate parameters with correlation and causal characteristics with the online monitoring data of the boiler pressure-bearing components are screened layer by layer, and the lag time of each valid candidate parameter is determined. The effective candidate parameters are selected by using the Lasso regression model to extract core influencing parameters, quantify the influence weight of a single core influencing parameter, and construct a quantitative correlation matrix of operating parameters and pressure component health status by combining the lag time. Based on the influence weight and lag time of each core influencing parameter in the quantitative correlation matrix of operating parameters and pressure component health status, a deep peak shaving optimization control scheme for generator sets is output.

[0007] In one possible implementation of the first aspect, the boiler operating parameter time series data includes at least one of the following: unit load, furnace outlet flue gas temperature, furnace negative pressure, flue gas oxygen content, primary air volume, secondary air volume, fuel coal feed rate, feedwater flow rate, main steam temperature, main steam pressure, and coal mill start-up and shutdown operation status parameters. The boiler pressure-bearing components include at least one of water-cooled walls, superheaters, and reheaters, and the corresponding online monitoring data include at least one of measuring point wall temperature, pipe wall expansion, structural strain stress, local deformation, and real-time fatigue damage data. In one possible implementation of the first aspect, the collected time-series data of boiler operating parameters and the online monitoring data of boiler pressure-bearing components undergo time alignment, anomaly removal, missing value imputation, and noise reduction normalization preprocessing, including: An interpolation algorithm is used to unify heterogeneous data with different sampling frequencies to the same timestamp node; The 3σ criterion was used to identify and remove abruptly abnormal measurement point data, and linear interpolation combined with interpolation of the mean of neighboring samples was used to fill in short-term missing data. Random interference noise is filtered out by wavelet threshold filtering combined with moving average filtering, and all preprocessed parameters are normalized to the [0, 1] interval.

[0008] In one possible implementation of the first aspect, the relevance preselection includes: Pearson correlation coefficient was used to screen parameters that showed a linear correlation with the online monitoring data of the boiler's pressure-bearing components; The mutual information method based on kernel density estimation to solve the joint probability is used to screen parameters that have a nonlinear correlation with the online monitoring data of the boiler pressure-bearing components; When the Pearson correlation coefficient or mutual information value corresponding to any parameter exceeds the preset association threshold, the parameter is retained and added to the candidate set.

[0009] In one possible implementation of the first aspect, the temporal causality check includes: Granger causality test based on vector autoregression (VAR) model and F test is used to determine the linear time-series causal relationship between operating parameters and the state of pressure-bearing components. The strength of nonlinear causality between operating parameters and the state of pressure-bearing components is calculated by transferring entropy. Remove spurious correlation parameters that have no actual causal relationship and retain parameters that meet the causal conditions.

[0010] In one possible implementation of the first aspect, the dynamic time warping similarity matching includes: The optimal matching cost between the parameter sequence and the component state sequence is calculated using a dynamic time warping algorithm. The lag time corresponding to each parameter is then corrected and determined based on the minimum matching cost.

[0011] In one possible implementation of the first aspect, a Lasso regression model is used to perform feature screening of the effective candidate parameters and quantify the influence weight of individual core influence parameters, including: Using online monitoring data of boiler pressure-bearing components as the dependent variable and the selected effective candidate parameters as independent variables, a Lasso regression objective function with an L1 regularization term is constructed. The L1 regularization term is used to compress the coefficients of irrelevant features to zero, thereby filtering out the core influencing parameters; The proportion of the regression coefficient of a single core influence parameter to the sum of the absolute values ​​of the regression coefficients of all core influence parameters is used as the influence weight of that single core influence parameter.

[0012] In one possible implementation of the first aspect, the deep peak-shaving optimization control scheme for the output generator set includes: Based on the influence weight of each core influencing parameter, the fluctuation amplitude limit of highly sensitive operating parameters is set, and the advanced control logic is configured in combination with the lag time. Deep peak shaving optimization instructions are generated from at least one dimension of unit load change rate control, furnace air distribution optimization, desuperheating water switching control and combustion organization adjustment, and the deep peak shaving optimization instructions are input to the unit load optimization control system.

[0013] According to a second aspect of the present invention, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the aforementioned generator set time-domain correlation analysis method based on online monitoring.

[0014] According to a third aspect of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing a computer program, which, when executed by a processor, implements the aforementioned method for time-domain correlation analysis of generator sets based on online monitoring.

[0015] According to a fourth aspect of the present invention, a computer program product is provided, which, when executed by a processor, implements the aforementioned method for time-domain correlation analysis of generator sets based on online monitoring.

[0016] Compared with the prior art, the present invention has at least the following beneficial effects: This invention synchronously collects multi-source data based on the clock sampling time of a distributed control system (DCS) and performs time alignment, noise reduction, and normalization preprocessing, breaking the traditional isolation between two sets of monitoring data and constructing a standardized time-series dataset for coupled analysis. On this basis, a hierarchical progressive screening architecture of correlation pre-selection and time-series causality testing is introduced, effectively eliminating accidental correlations in massive operational data and accurately locating the real physical disturbance source that induces damage to pressure-bearing components. Simultaneously, by using dynamic time warping similarity matching, the lag time of parameter fluctuation transmission is accurately calculated, overcoming the technical deficiency of traditional methods in capturing the heat conduction time difference pattern under deep peak-shaving conditions. Subsequently, a Lasso regression model is used for deep feature screening and influence weight calculation, completely eliminating the blindness of relying on subjective parameter selection based on human experience. Finally, based on a quantitative correlation matrix containing influence weights and lag time, an optimized control scheme is output to guide the system to reasonably constrain highly sensitive parameters and configure advanced control logic, directly transforming the algorithm's time-domain analysis results into deep peak-shaving operation guidance. This fundamentally stabilizes the wall temperature fluctuation of the heated surface, significantly reduces the amplitude of alternating thermal stress in components, and effectively delays the accumulation of pipe wall fatigue degradation. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the specific embodiments of the present invention, the drawings used in the description of the specific embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0018] Figure 1 This is a flowchart of a generator set time-domain correlation analysis method based on online monitoring according to the present invention.

[0019] Figure 2 This is a flowchart illustrating a time-domain correlation analysis method for generator sets based on online monitoring, as an example. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] The technical problem this invention aims to solve is that under the normalized deep peak-shaving operation of thermal power units, the large fluctuations in boiler operating parameters cause the boiler's pressure-bearing and heat-receiving components to be in an alternating temperature field for a long time, inducing repeated alternating thermal stress and exacerbating thermal fatigue degradation. Existing technologies use two sets of isolated monitoring data and lack systematic time-domain correlation analysis methods, making it impossible to effectively distinguish between accidental data correlations and real causal disturbances, and also unable to provide precise quantitative guidance for peak-shaving control. The core idea of ​​this invention is to utilize the DCS data interface of the generator unit and sensor hardware to access multi-source monitoring data. Based on preprocessing methods such as time alignment and noise reduction normalization to eliminate the influence of heterogeneous dimensions, a hierarchical progressive time-domain analysis architecture is constructed, consisting of correlation pre-selection, time-series causality verification, and dynamic time warping and time-series matching. Redundant and pseudo-correlated variables are eliminated layer by layer, and the lag time introduced by heat conduction and other factors is locked. Finally, a Lasso regression model with L1 regularization is used to quantitatively extract core influencing parameters and evaluate their weights, outputting a deep peak-shaving optimization control scheme with advanced control logic.

[0022] The following is combined Figure 1 and Figure 2 The specific steps of this invention will be described in detail below. This invention provides a method for time-domain correlation analysis of generator sets based on online monitoring, specifically including the following steps: S1. Based on the unified sampling time of the generator set's distributed control system (DCS), synchronously collect boiler operating parameter timing data and online monitoring data of boiler pressure-bearing components.

[0023] Specifically, the entire analysis system is deployed and operates within the generator unit's online monitoring platform. The multi-source data acquisition module connects to both the boiler operation measurement point link and the pressure-bearing component sensor acquisition link, directly acquiring existing DCS data from the power plant and data from the distributed sensor acquisition hardware. To achieve cross-system joint analysis, the multi-source data acquisition module uses the unit's DCS system clock as a globally unified reference for time calibration, pre-sets fixed sampling intervals, and freely extracts continuous time-series datasets within any specified analysis period according to control requirements.

[0024] The boiler operating parameter time-series data includes at least one of the following: unit load, furnace outlet flue gas temperature, furnace negative pressure, flue gas oxygen content, primary air volume, secondary air volume, fuel and coal feed rate, feedwater flow rate, main steam temperature, main steam pressure, and coal mill start-up and shutdown status parameters. The boiler pressure-bearing components include at least one of the boiler's water-cooled walls, superheaters, and reheaters, and their corresponding online monitoring data includes at least one of the following: measuring point wall temperature, tube wall expansion, structural strain and stress, local deformation, and real-time fatigue damage data. All collected raw continuous time-series data are directly cached in the time-series database built into the unit's online monitoring platform.

[0025] S2. Perform time alignment, anomaly removal, missing value filling, and noise reduction normalization preprocessing on the collected boiler operating parameter time series data and boiler pressure component online monitoring data to generate a standardized time series dataset.

[0026] Specifically, the time alignment unit in the time series data preprocessing module first uses an interpolation algorithm to map asynchronous heterogeneous data sampled at different sampling frequencies due to different field links to the same timestamp node, so that all subsequent monitoring parameters to be processed fall on the same discrete time series node.

[0027] Subsequently, the anomaly repair unit uses the 3σ criterion to perform sliding discrimination on the time-aligned time-series waveform, automatically identifying and removing abruptly abnormal measurement point data introduced by field environmental interference, instantaneous jumps in sensor hardware, etc.; for the missing positions generated after data removal or the short-term data gaps existing in the original acquisition link itself, a combination of linear interpolation and nearest sample mean interpolation is used to complete the missing value filling, so as to ensure the continuity of the data sequence.

[0028] Finally, the noise reduction unit uses a wavelet threshold filtering algorithm combined with a moving average filtering algorithm to remove high-frequency random interference noise from the data. After preprocessing, all parameters that have been noise-reduced and repaired are normalized to map parameters of different physical dimensions and ranges to the [0, 1] interval, thus generating a multi-dimensional, standardized time series dataset.

[0029] S3. For the standardized time series dataset, perform correlation pre-selection, time series causality test and dynamic time warping similarity matching in sequence to screen out effective candidate parameters that have correlation and causal characteristics with the online monitoring data of boiler pressure components, and determine the effect lag time of each effective candidate parameter.

[0030] Specifically, the hierarchical time-domain correlation analysis module first performs preliminary screening of the first layer of operating parameters through a correlation pre-selection unit. In this unit, Pearson correlation coefficient is used to screen linear correlation parameters between operating parameters and online monitoring data of boiler pressure components, while mutual information method based on kernel density estimation to solve joint probabilities is used to screen nonlinear correlation parameters between the two sets of data.

[0031] The formula for calculating the Pearson correlation coefficient is as follows:

[0032] In the formula, The calculated Pearson correlation coefficient; For a certain boiler operating parameter at discrete time... Standardized values, This is the average value of the operating parameter over the analysis period; For the discrete time intervals of online monitoring data of a certain boiler pressure-bearing component Standardized values, This represents the average value of the monitoring data for this pressure-bearing component.

[0033] The formula for calculating mutual information is as follows:

[0034] In the formula, For running parameters Monitoring data of pressure-bearing components The nonlinear mutual information value between them; This is the joint probability density distribution obtained by the kernel density estimation method; and These are the corresponding marginal probability density distributions.

[0035] During the screening process, the initial screening is completed according to the system's preset correlation threshold. When the absolute value of the Pearson correlation coefficient or the mutual information value corresponding to any parameter exceeds the preset correlation threshold, the parameter is retained and added to the candidate set.

[0036] The parameters selected through the first preliminary layer are then input into the time-series causality testing unit for the second layer of screening. The time-series causality testing unit uses a vector autoregressive time-series model combined with the F-test to perform a Granger causality test to determine linear time-series causal relationships. The model expression it relies on is as follows:

[0037] In the formula, This represents the current state value of the pressure-bearing component. and The time lag order is respectively Historical standardized values ​​of operating parameters and pressure-bearing component conditions at the time of operation; and These are the regression coefficients of the model; This is the residual term.

[0038] Meanwhile, the strength of the nonlinear causal relationship between operating parameters and the state of pressure-bearing components is calculated using the transfer entropy. The formula for calculating the transfer entropy is as follows:

[0039] In the formula, The numerical value for nonlinear causality; This represents the information entropy of the corresponding joint variable; It represents the conditional entropy under given historical variables; the larger the value, the higher the nonlinear causal contribution.

[0040] Through a dual-track causality test, spuriously correlated parameters with no actual causal relationship are eliminated, retaining only those that meet the causal conditions. The remaining valid parameters are imported into the dynamic time warping similarity matching unit for third-level analysis. The optimal matching cost between the parameter sequence and the component state sequence is calculated using the dynamic time warping algorithm; the formulas for calculating the single-step and cumulative costs are as follows:

[0041]

[0042] In the formula, The distance between the operating parameter data points and the component status data points; This represents the cumulative matching cost. It is calculated based on the overall sequence matching error. The minimum matching result is used as the evaluation benchmark to correct and determine the lag time corresponding to each parameter.

[0043] S4. Use the Lasso regression model to screen the effective candidate parameters, extract the core influencing parameters, quantify the influence weight of a single core influencing parameter, and construct a quantitative correlation matrix of operating parameters and the health status of pressure-bearing components by combining the lag time.

[0044] Specifically, the online monitoring indicators of boiler pressure-bearing components are used as the dependent variable. The selected valid candidate parameters are used as independent variables. We construct the Lasso regression objective function with an L1 regularization term, and its formula is as follows:

[0045] In the formula, These are the regression coefficients; This represents the total number of sample points; The total number of valid candidate parameters for input; These are the preset regularization coefficients. The L1 regularization term is used to compress irrelevant feature coefficients to zero, automatically selecting the core influencing parameters.

[0046] Subsequently, the influence weight is calculated by using the proportion of the regression coefficient of a single core influence parameter to the sum of the absolute values ​​of the regression coefficients of all core influence parameters, as shown in the following formula:

[0047] In the formula, The influence weight of the core influencing parameters; represents the corresponding regression coefficient.

[0048] Finally, the influence weights, lag times, and correlation attributes (linear or nonlinear) of each parameter are summarized to generate a quantitative correlation matrix, which is then stored locally.

[0049] S5. Based on the influence weight and lag time of each core influencing parameter in the correlation matrix of operating parameters and health status of pressure components, output the deep peak shaving optimization control scheme of the generator set.

[0050] Specifically, the peak-shaving strategy optimization output module reads the quantified correlation matrix data and constrains the fluctuation amplitude of highly sensitive operating parameters based on the influence weight of each core influencing parameter. Simultaneously, it implements advanced control logic configuration by incorporating the lag time, introducing control variables in advance, and generating standardized deep peak-shaving optimization instructions from four dimensions: unit load change rate control, furnace air distribution optimization, desuperheating water switching control, and combustion organization adjustment. The optimization results are directly input to the unit load optimization control system for online guidance. During actual system operation, the threshold values, filtering orders, and regularization coefficients of each algorithm are adjusted. The parameter correction coefficients can be fine-tuned online based on the unit's installed capacity, boiler type, commonly used coal type, and annual peak load range. The fine-tuning range is constrained by the boiler's design limits and the unit's actual operating procedures.

[0051] This invention breaks the traditional isolation between two sets of monitoring data by using unified sampling based on the DCS clock, combined with interpolation and joint filtering normalization. This eliminates the dimensional differences and random interference of multi-source heterogeneous time-series data. By constructing a hierarchical progressive time-domain analysis architecture based on correlation pre-selection, Granger causality and transfer entropy testing, and dynamic time warping, it can peel away the objectively existing random data correlations in massive operating data layer by layer under the non-stationary characteristics of generator unit deep peak shaving normal operation. It can accurately locate effective parameters with real physical mechanism correlations and accurately correct and determine the lag time caused by physical heat transfer delay. This overcomes the difficulty of adapting to time lag and the difficulty of identification in existing technologies. The limitations of relying solely on subjective selection based on human experience were overcome by introducing a Lasso regression objective function and coefficient ratio calculation. This approach enabled data-driven dimensionality reduction extraction and objective quantitative evaluation of core influencing parameters, overcoming the limitations of subjective selection based on human experience. It directly reflects the contribution ratio of various operating parameters to the thermal fatigue damage of pressure-bearing components. Finally, by constraining highly sensitive parameters through a quantitative correlation matrix and configuring advanced control logic, it directly guided the refined control processes such as unit load rate management, air distribution ratio, and desuperheating water switching. This reduced the alternating thermal stress amplitude of pressure-bearing components under deep peak-shaving disturbances from the data source, effectively slowing down the fatigue degradation accumulation rate of pipe wall materials and improving the safety and economic benefits of operation over a wide load range.

[0052] In another embodiment of the present invention, a computer device is provided, comprising a processor and a memory. The memory stores a computer program, which includes program instructions. The processor executes the program instructions stored in the computer storage medium. The processor may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing and control core of the terminal, suitable for implementing one or more instructions, specifically suitable for loading and executing one or more instructions from the computer storage medium to achieve a corresponding method flow or corresponding function. The processor described in this embodiment of the present invention can be used for the operation of a generator set time-domain correlation analysis method based on online monitoring.

[0053] In another embodiment of the present invention, a storage medium is provided, specifically a computer-readable storage medium (Memory), which is a memory device in a computer device used to store programs and data. It is understood that the computer-readable storage medium here can include both the built-in storage medium in the computer device and extended storage media supported by the computer device. The computer-readable storage medium provides storage space that stores the terminal's operating system. Furthermore, the storage space also stores one or more instructions suitable for loading and execution by a processor. These instructions can be one or more computer programs (including program code). It should be noted that the computer-readable storage medium here can be Random Access Memory (RAM) or non-volatile memory, such as at least one disk storage device. The processor can load and execute one or more instructions stored in the computer-readable storage medium to implement the corresponding steps of the generator set time-domain correlation analysis method based on online monitoring in the above embodiments.

[0054] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, optical storage, etc.) containing computer-usable program code.

[0055] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0056] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0057] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0058] This invention also provides a computer program product for executing any of the above-described online monitoring-based generator set time-domain correlation analysis methods. Since the computer program product provided by this invention belongs to the same inventive concept as the online monitoring-based generator set time-domain correlation analysis method described above, it possesses all the advantages of the online monitoring-based generator set time-domain correlation analysis method described above. Therefore, the beneficial effects of the computer program product provided by this invention will not be elaborated upon here.

[0059] In this invention, the terms "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to a specific feature, structure, material, or characteristic described in connection with that embodiment or example, which is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0060] Finally, it should be noted that the above-described embodiments are merely specific implementations of the present invention, used to illustrate the technical solutions of the present invention, and not to limit them. The scope of protection of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments within the scope of the technology disclosed in the present invention, or make equivalent substitutions for some of the technical features; and these modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be covered within the scope of protection of the present invention.

Claims

1. A method for time-domain correlation analysis of generator sets based on online monitoring, characterized in that, include: Based on the unified sampling time of the generator set's distributed control system (DCS) clock, the timing data of boiler operating parameters and online monitoring data of boiler pressure-bearing components are collected synchronously. The collected time-series data of boiler operating parameters and the online monitoring data of boiler pressure-bearing components are preprocessed by time alignment, anomaly removal, missing value filling and noise reduction normalization to generate a standardized time-series dataset. The standardized time-series dataset is subjected to correlation pre-selection, time-series causality test and dynamic time warping similarity matching in sequence. Valid candidate parameters with correlation and causal characteristics with the online monitoring data of the boiler pressure-bearing components are screened layer by layer, and the lag time of each valid candidate parameter is determined. The effective candidate parameters are selected by using the Lasso regression model to extract core influencing parameters, quantify the influence weight of a single core influencing parameter, and construct a quantitative correlation matrix of operating parameters and pressure component health status by combining the lag time. Based on the influence weight and lag time of each core influencing parameter in the quantitative correlation matrix of operating parameters and pressure component health status, a deep peak shaving optimization control scheme for generator sets is output.

2. The method for time-domain correlation analysis of generator sets based on online monitoring according to claim 1, characterized in that, The boiler operating parameter time series data includes at least one of the following: unit load, furnace outlet flue gas temperature, furnace negative pressure, flue gas oxygen content, primary air volume, secondary air volume, fuel and coal feed rate, feedwater flow rate, main steam temperature, main steam pressure, and coal mill start-up and shutdown operation status parameters. The boiler pressure-bearing components include at least one of water-cooled walls, superheaters, and reheaters, and the corresponding online monitoring data include at least one of measuring point wall temperature, pipe wall expansion, structural strain stress, local deformation, and real-time fatigue damage data.

3. The method for time-domain correlation analysis of generator sets based on online monitoring according to claim 1, characterized in that, The collected time-series data of boiler operating parameters and the online monitoring data of boiler pressure-bearing components undergo time alignment, anomaly removal, missing value imputation, and noise reduction normalization preprocessing, including: An interpolation algorithm is used to unify heterogeneous data with different sampling frequencies to the same timestamp node; The 3σ criterion was used to identify and remove abruptly abnormal measurement point data, and linear interpolation combined with interpolation of the mean of neighboring samples was used to fill in short-term missing data. Random interference noise is filtered out by wavelet threshold filtering combined with moving average filtering, and all preprocessed parameters are normalized to the [0, 1] interval.

4. The method for time-domain correlation analysis of generator sets based on online monitoring according to claim 1, characterized in that, The correlation pre-selection includes: Pearson correlation coefficient was used to screen parameters that showed a linear correlation with the online monitoring data of the boiler's pressure-bearing components; The mutual information method based on kernel density estimation to solve the joint probability is used to screen parameters that have a nonlinear correlation with the online monitoring data of the boiler pressure-bearing components; When the Pearson correlation coefficient or mutual information value corresponding to any parameter exceeds the preset association threshold, the parameter is retained and added to the candidate set.

5. The method for time-domain correlation analysis of generator sets based on online monitoring according to claim 1, characterized in that, The temporal causality test includes: Granger causality test based on vector autoregression (VAR) model and F test is used to determine the linear time-series causal relationship between operating parameters and the state of pressure-bearing components. The strength of nonlinear causality between operating parameters and the state of pressure-bearing components is calculated by transferring entropy. Remove spurious correlation parameters that have no actual causal relationship and retain parameters that meet the causal conditions.

6. The method for time-domain correlation analysis of generator sets based on online monitoring according to claim 1, characterized in that, The dynamic time warping similarity matching includes: The optimal matching cost between the parameter sequence and the component state sequence is calculated using a dynamic time warping algorithm. The lag time corresponding to each parameter is then corrected and determined based on the minimum matching cost.

7. The method for time-domain correlation analysis of generator sets based on online monitoring according to claim 1, characterized in that, The effective candidate parameters are selected using a Lasso regression model, and the influence weight of each core influencing parameter is quantified, including: Using online monitoring data of boiler pressure-bearing components as the dependent variable and the selected effective candidate parameters as independent variables, a Lasso regression objective function with an L1 regularization term is constructed. The L1 regularization term is used to compress the coefficients of irrelevant features to zero, thereby filtering out the core influencing parameters; The proportion of the regression coefficient of a single core influence parameter to the sum of the absolute values ​​of the regression coefficients of all core influence parameters is used as the influence weight of that single core influence parameter.

8. The method for time-domain correlation analysis of generator sets based on online monitoring according to claim 1, characterized in that, The deep peak-shaving optimization control scheme for the output generator set includes: Based on the influence weight of each core influencing parameter, the fluctuation amplitude limit of highly sensitive operating parameters is set, and the advanced control logic is configured in combination with the lag time. Deep peak shaving optimization instructions are generated from at least one dimension of unit load change rate control, furnace air distribution optimization, desuperheating water switching control and combustion organization adjustment, and the deep peak shaving optimization instructions are input to the unit load optimization control system.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the generator set time-domain correlation analysis method based on online monitoring as described in any one of claims 1 to 8.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the generator set time-domain correlation analysis method based on online monitoring as described in any one of claims 1 to 8.

Images