Coal feeder coal breaking fault diagnosis method and device based on improved pelican algorithm

Abstract

According to the coal feeder coal breaking fault diagnosis method and device based on the improved pelican algorithm provided by the embodiment of the invention, the initial center of the fuzzy C-means clustering is accurately optimized by introducing the improved pelican optimization algorithm, so that dynamic adjustment and adaptive learning of a fault diagnosis model are realized; the core defects that a traditional method is insufficient in sample data and prone to falling into local optimum are effectively overcome. According to the method, full-process automation from data acquisition, feature extraction and clustering analysis to state recognition and early warning is realized, the accuracy and the real-time performance of diagnosis are remarkably improved, and meanwhile, the robustness and the self-adaptive capability of the system in a complex industrial environment are enhanced. In the operation scene of the coal feeder, the method can efficiently identify the coal interruption fault, realize early warning and optimize the maintenance strategy, so that the operation efficiency and reliability of equipment are improved, the operation and maintenance cost is reduced, and the method has important engineering application value.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method and apparatus for diagnosing coal feeder failure based on an improved Pelican algorithm. Background Technology

[0002] As a core component ensuring boiler combustion efficiency, the operational stability of the fuel system in thermal power plants directly impacts the overall performance of the unit. With the acceleration of industrial intelligence, data-driven fault diagnosis technology is gradually becoming an important direction for fuel system monitoring. Specifically, this technology system covers the entire process from raw coal transportation and grinding to combustion control, including key aspects such as DCS data acquisition, multi-parameter dynamic modeling, and unsupervised cluster analysis. Among these, the coal feeder, as a critical execution unit for fuel supply, can cause a chain reaction of idling in the coal mill, abnormal coal powder preparation, and a sharp drop in combustion efficiency if it experiences a coal shortage. Existing technologies mostly rely on manual experience or traditional machine learning methods, but these suffer from fundamental defects such as strong sample dependence and poor algorithm convergence.

[0003] However, existing fault diagnosis methods directly employ fixed thresholds or single clustering algorithms without considering the multidimensional dynamic characteristics of equipment operating parameters. This can lead to a high false alarm rate (>15%) or a response lag (>30 minutes). Specifically, the traditional FCM algorithm is sensitive to the initial cluster centers, and its objective function... During the iteration process, it is prone to getting trapped in local optima, and in industrial scenarios with insufficient sample data (fault samples <5%), it is difficult to guarantee diagnostic accuracy. In addition, existing systems lack the ability to identify transitional states and cannot establish an early warning mechanism before a fault occurs, leading to frequent occurrences of neglected or excessive maintenance strategies, resulting in serious consequences such as losses of tens of thousands of yuan per hour for a single downtime. Summary of the Invention

[0004] The core idea of ​​this invention is to achieve comprehensive perception and feature extraction of the coal feeder's operating status by constructing a dynamic feature space that integrates multi-dimensional operating parameters such as coal feeder speed, current, coal mill related flow rate, differential pressure, temperature, and air pressure. This method first collects and normalizes operating data through a DCS system. Then, it employs an improved pelican optimization algorithm, using a population initialization strategy that simulates pelican predation behavior to accurately optimize the initial centers of fuzzy C-means clustering. Fitness is defined by a clustering objective function, and iterative optimization is performed, effectively overcoming the shortcomings of traditional methods that are sensitive to initial values ​​and prone to getting trapped in local optima. Based on the optimized cluster centers, iterative fuzzy C-means clustering is performed. By calculating the Euclidean distance between the monitored features and the standard cluster centers, the coal feeder's operating status is accurately classified into four categories: normal, coal supply interruption transition, coal supply interruption fault, and abnormal. When a transition or fault state is detected, the system automatically triggers an audible and visual alarm. For abnormal states that exceed the preset threshold, the model parameters are updated synchronously, thus forming a closed-loop fault diagnosis process from state perception, intelligent diagnosis to early warning feedback, which significantly improves the system's diagnostic accuracy, adaptability and real-time performance in complex industrial scenarios.

[0005] To achieve the above objectives, this invention provides a method for diagnosing coal feeder failures based on an improved pelican algorithm. The method includes: S1, selecting multi-dimensional operating parameters including feeder speed, feeder current, mill current, mill primary air flow rate, mill inlet and outlet differential pressure, mill outlet temperature, mill outlet air-coal pressure, and mill inlet primary air pressure as model input features to construct a dynamic feature space; S2, collecting operating data during coal feeder failures through a DCS system, dividing the data into three stages: normal, transition, and failure, and performing normalization processing; S3, using an improved pelican optimization algorithm to optimize the initial centers of fuzzy C-means clustering, including initializing the number of clusters and IP. OA parameters and FCM parameters are optimized by using a population initialization strategy that simulates pelican predation behavior to optimize the initial cluster centers. The fitness function is defined by the FCM objective function and particle iteration optimization is performed. S4, FCM clustering iteration is performed based on the optimized initial cluster centers. The operating status of the coal feeder is judged according to the Euclidean distance between the monitored feature parameters and the standard class center point. The status is divided into four categories: normal (1), coal interruption transition (2), coal interruption fault (3), and abnormal (0). S5, when the diagnosis result is a coal interruption transition state or a coal interruption fault state, the audible and visual alarm module is triggered to generate alarm information. If the Euclidean distance between the monitored feature parameters and the standard class center point exceeds the preset threshold, it is classified as an abnormal state and the model parameters are updated.

[0006] Furthermore, the improved pelican optimization algorithm for optimizing the initial centers of fuzzy C-means clustering also includes: S31, generating initial cluster centers through a population initialization strategy simulating pelican predation behavior, specifically including: randomly initializing the population position vector. ,in , For population size; according to Update individual location, where This is the step size coefficient. The current optimal solution; S32, with the FCM objective function Define the fitness function, where For fuzzy membership, As cluster center, It is a fuzzy weighted index.

[0007] Furthermore, the FCM clustering iteration based on the optimized initial cluster centers also includes: S41, calculating the monitoring feature parameters. Center point of standard class European distance ,in , S42, based on the preset threshold Classify, when When the condition is determined to be abnormal (0), it is considered an abnormal state.

[0008] Furthermore, when the diagnosis result is a coal shortage transition state or a coal shortage fault state, the audible and visual alarm module also includes: S51, using an LED light group and a buzzer to achieve an audible and visual alarm, wherein the LED light group uses red and green dual-color indicator lights, with a solid red light indicating a coal shortage fault state (3), and a flashing red light indicating a coal shortage transition state (2); S52, according to Calculate signal transmission delay to ensure real-time alarm information, where For transmission power, For transmission distance, For bandwidth, This refers to the signal-to-noise ratio.

[0009] Furthermore, the operational data collected during the coal shortage incident through the DCS system also includes: S21, using a sliding window method to segment the collected data, with a window length of [missing information]. Each sampling period, with an overlap rate of [number] times. S22, Perform wavelet denoising on the normalized data, retaining... The signal components.

[0010] Furthermore, a method for diagnosing coal feeder failures based on an improved pelican algorithm also includes: S6, dynamically adjusting the Euclidean distance threshold based on historical operating data. Specifically, this includes: calculating the mean Euclidean distance of normal state samples within the current period. and standard deviation ;according to Update threshold, where , .

[0011] In another aspect, this invention provides a coal feeder coal shortage fault diagnosis device based on an improved pelican algorithm. The device includes: a multi-dimensional parameter input module, used to select multi-dimensional operating parameters including coal feeder speed, coal feeder current, pulverizer current, pulverizer primary air flow rate, pulverizer inlet and outlet differential pressure, pulverizer outlet temperature, pulverizer outlet air-coal pressure, and pulverizer inlet primary air pressure as model input features to construct a dynamic feature space; a data acquisition and preprocessing module, used to acquire operating data during a coal shortage incident through a DCS system, divide the data into three stages—normal, transition, and fault—and perform normalization processing; and a cluster center optimization module, used to optimize the initial centers of fuzzy C-means clustering using an improved pelican optimization algorithm, including initializing the number of clusters, IPOA parameters, and FCM parameters, optimizing the initial cluster centers through a population initialization strategy simulating pelican predation behavior, with the FCM objective function... Define the fitness function and perform particle iterative optimization; the state classification module is used to perform FCM clustering iteration based on the optimized initial cluster centers, according to the Euclidean distance between the monitored feature parameters and the standard cluster centers. The operating status of the coal feeder is determined and classified into four categories: normal (1), coal interruption transition (2), coal interruption fault (3), and abnormal (0). The alarm and parameter update module is used to trigger the audible and visual alarm module to generate alarm information when the diagnosis result is a coal interruption transition state or a coal interruption fault state. If the Euclidean distance between the monitored characteristic parameter and the center point of the standard class exceeds the preset threshold, the alarm will be updated accordingly. If it does, it is classified as an abnormal state and the model parameters are updated.

[0012] Furthermore, the cluster center optimization module is also used to: generate initial cluster centers through a population initialization strategy simulating pelican predation behavior, specifically including: randomly initializing the population position vector. ,in , For population size; according to Update individual location, where This is the step size coefficient. This is the current optimal solution; using the FCM objective function Define the fitness function and perform particle iterative optimization, where For fuzzy membership, As cluster center, It is a fuzzy weighted index.

[0013] Furthermore, the state classification module is also used to: calculate monitoring feature parameters. Center point of standard class European distance ,in , According to the preset threshold Classify, when When the condition is determined to be abnormal (0), it is considered to be in an abnormal state.

[0014] Furthermore, the alarm and parameter update module is also used to: realize audible and visual alarms through a combination of LED lights and a buzzer, wherein the LED lights use red and green dual-color indicator lights, with a solid red light indicating a coal shortage fault state (3), and a flashing red light indicating a coal shortage transition state (2); according to Calculate signal transmission delay to ensure real-time alarm information, where For transmission power, For transmission distance, For bandwidth, This refers to the signal-to-noise ratio.

[0015] Furthermore, the data acquisition and preprocessing module is also used to: segment the acquired data using a sliding window method, with a window length of [missing information]. Each sampling period, with an overlap rate of [number] times. Perform wavelet denoising on the normalized data, retaining... The signal components.

[0016] Furthermore, a coal feeder coal shortage fault diagnosis device based on the improved pelican algorithm also includes: a dynamic threshold adjustment module, used to dynamically adjust the Euclidean distance threshold according to historical operating data. Specifically, this includes: calculating the mean Euclidean distance of normal state samples within the current period. and standard deviation ;according to Update threshold, where , .

[0017] Compared with the prior art, the technical solution provided by the present invention has at least the following beneficial effects:

[0018] 1. Deep integration of multi-dimensional operating parameters: By constructing a dynamic feature space that includes multi-dimensional characteristic parameters such as feeder speed, current and coal mill related flow, differential pressure, temperature and pressure, the operating status of the feeder is comprehensively characterized, providing a rich data foundation for accurate diagnosis.

[0019] 2. Intelligent clustering optimization mechanism: The improved pelican optimization algorithm is used to optimize the initial centers of fuzzy C-means clustering. By simulating the population initialization strategy of pelican predation behavior, the shortcomings of traditional methods, such as sensitivity to initial values ​​and easy getting trapped in local optima, are effectively overcome, and the clustering quality and diagnostic reliability are significantly improved.

[0020] 3. Real-time status identification and early warning: Based on the optimized cluster centers, the system calculates the Euclidean distance between the monitoring features and the standard cluster centers to achieve a fine division of the coal feeder's operating status. This enables timely identification of coal shortage transitions and fault states and triggers alarms. Simultaneously, through a parameter self-updating mechanism under abnormal conditions, a complete diagnosis-early warning-optimization closed loop is formed, significantly improving the system's real-time performance and adaptability in industrial field applications.

[0021] This invention discloses a method and apparatus for diagnosing coal feeder failures based on an improved pelican algorithm. By constructing a multi-dimensional dynamic feature space of operating parameters and employing the improved pelican optimization algorithm to optimize the initial centers of fuzzy C-means clustering, it effectively overcomes the core defects of traditional methods, such as sensitivity to initial values ​​and susceptibility to local optima. The apparatus achieves full automation from data acquisition, feature extraction, cluster optimization to state identification and early warning, significantly improving the accuracy of fault diagnosis and the system's adaptive capability. In engineering applications, it can promptly identify coal feeder transitions and fault states and trigger early warnings. Simultaneously, through a parameter self-updating mechanism under abnormal conditions, a complete diagnostic optimization closed loop is formed, thereby significantly improving the reliability of coal feeder operation and effectively reducing maintenance costs. Attached Figure Description

[0022] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 A flowchart illustrating a coal feeder coal shortage fault diagnosis method based on an improved pelican algorithm, provided in an embodiment of the present invention; Figure 2 A flowchart of an alarm system for diagnosing coal feeder failure based on an improved pelican algorithm, provided as an embodiment of the present invention; Figure 3 A flowchart of the judgment state for coal feeder coal shortage fault diagnosis based on the improved Pelican algorithm is provided in an embodiment of the present invention. Figure 4 A schematic diagram of the structure of an alarm system for diagnosing coal feeder failure based on an improved pelican algorithm, provided in an embodiment of the present invention; Figure 5 This is a schematic diagram of a coal feeder coal shortage fault diagnosis device based on the improved Pelican algorithm, provided in an embodiment of the present invention. Detailed Implementation

[0023] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0024] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0025] The core idea of ​​this invention is to achieve comprehensive perception of equipment operating status by constructing a dynamic feature space that integrates multi-dimensional operating parameters such as coal feeder speed, current, and coal mill flow rate, differential pressure, temperature, and pressure. Based on this, an improved pelican optimization algorithm is adopted. By simulating the pelican's predatory behavior and employing a population initialization strategy, the initial centers of fuzzy C-means clustering are precisely optimized, effectively overcoming the core defects of traditional methods that are sensitive to initial values ​​and prone to getting trapped in local optima. Based on the optimized cluster centers, the system calculates the Euclidean distance between the monitored features and the standard cluster centers to achieve a fine division of the coal feeder's operating status, enabling timely identification of coal supply interruption transitions and fault states, and triggering audible and visual alarms. By introducing a dynamic threshold adjustment mechanism and parameter self-updating function, a complete optimization closed loop from status perception and intelligent diagnosis to early warning feedback is formed, significantly improving the accuracy of fault diagnosis and the system's adaptability in industrial field applications. Ultimately, while ensuring the stable operation of the coal feeder, it achieves a comprehensive improvement in operation and maintenance efficiency and effective cost control.

[0026] The following description, with reference to the accompanying drawings, describes a method and apparatus for diagnosing coal feeder failure based on an improved pelican algorithm, according to an embodiment of the present invention.

[0027] Example 1 This embodiment provides a method for diagnosing coal feeder failures based on an improved pelican algorithm. For example... Figure 1 As shown, the method includes the following steps: S1. Select multi-dimensional operating parameters, including coal feeder speed, coal feeder current, coal mill current, coal mill primary air flow rate, coal mill inlet and outlet differential pressure, coal mill outlet temperature, coal mill outlet air-coal pressure and coal mill inlet primary air pressure, as model input features to construct a dynamic feature space.

[0028] Specifically, in the technical solution of this invention, step one, "selecting multi-dimensional operating parameters including coal feeder speed, coal feeder current, pulverizer current, pulverizer primary air flow rate, pulverizer inlet and outlet differential pressure, pulverizer outlet temperature, pulverizer outlet air-coal pressure, and pulverizer inlet primary air pressure as model input features and constructing a dynamic feature space," is a key step in realizing the structuring and feature engineering of the input data for the fault diagnosis model. This step systematically filters operating parameters highly correlated with coal feeder failures to construct a multi-dimensional feature space with physical meaning and dynamic response capabilities, providing a high-quality data foundation for subsequent cluster analysis and fault identification.

[0029] Specifically, this step first involves performing correlation analysis and feature extraction on the operating parameters of several key devices in the fuel system based on industrial process control theory and historical operating data. Specifically, the selected features include: coal feeder speed (…). ), coal feeder current ( ), coal mill current ( ), primary air flow rate of coal mill ( ), coal mill inlet and outlet differential pressure ( ), coal mill outlet temperature ( ), coal mill outlet air-coal pressure ( ) and the primary air pressure at the inlet of the coal mill ( These parameters reflect the coal feeding capacity of the coal feeder, the motor load status, the operating conditions of the coal mill, and the gas-solid two-phase flow characteristics, and have strong physical interpretability.

[0030] Furthermore, at the parameter level, the acquisition frequency of each parameter should be no less than 1Hz to ensure the ability to capture dynamic features. Normalization is performed using the Min-Max Normalization method, mapping the original data to the [0,1] interval, as shown in the formula:

[0031] in, These are the original parameter values. and These are the minimum and maximum values ​​of the parameter in historical data, respectively. After normalization, the parameters are comparable, which is beneficial for the stable operation of subsequent clustering algorithms.

[0032] Specifically, this step applies to real-time monitoring and fault early warning of coal feeders and coal mills in the fuel system of thermal power plants. The above parameters are collected through a DCS system, and data preprocessing includes cleaning, missing value imputation, and outlier removal to ensure the integrity and reliability of the input data. The constructed dynamic feature space can be used as input to the IPOA-FCM clustering model to identify subtle changes in operating status, thereby achieving early warning of coal shortage faults.

[0033] Specifically, this step, through the appropriate selection and normalization of multidimensional parameters, effectively improves the model's sensitivity to and discriminative ability regarding operational states. The constructed dynamic feature space not only reflects the operational differences of equipment under normal, transitional, and fault states, but also provides solid data support for subsequent optimization of cluster centers and improvement of diagnostic accuracy. This method maintains high clustering stability even with limited sample data, laying the foundation for realizing an intelligent and automated fault diagnosis system.

[0034] S2 collects operational data during coal shortage accidents through the DCS system, divides the data into three stages: normal, transition, and fault, and performs normalization processing.

[0035] Specifically, in the technical solution of this invention, step two, "collecting operational data during a coal shortage accident through a DCS system, dividing the data into three stages—normal, transition, and fault—and performing normalization processing," is a crucial preliminary step for training the fault diagnosis model and identifying its state. This step is technically implemented based on the real-time data acquisition capabilities of the Distributed Control System (DCS), combined with the temporal characteristics of industrial process state evolution, to structurally divide and standardize the raw data, thereby improving the convergence efficiency and diagnostic accuracy of subsequent clustering algorithms.

[0036] Specifically, the DCS system collects multi-dimensional operating parameters during coal shortage accidents through its engineering workstation or data interface, including but not limited to feeder speed, feeder current, pulverizer current, primary air flow rate, inlet and outlet differential pressure, outlet temperature, outlet air-coal pressure, and inlet primary air pressure. The acquisition frequency is typically set to 1 to 10 times per second to ensure the temporal continuity and dynamic response capability of the data. The collected data is stored in time series format for easy subsequent stage division.

[0037] Furthermore, based on the evolution of the coal shortage fault, the data is divided into three stages: Normal, Transition, and Fault. The Normal stage refers to the time window when equipment operates stably without abnormal signals; the Transition stage is the intermediate state where a coal shortage fault is about to occur and operating parameters begin to deviate from the normal range; the Fault stage is the period when a coal shortage has already occurred and system operating parameters are significantly abnormal. The classification criteria can be based on timestamps, the time preceding the fault occurrence (e.g., 10 minutes prior), and parameter change rate thresholds (e.g., current change rate exceeding 0.5 A / s).

[0038] Specifically, the normalization process employs the Min-Max Normalization method, mapping each parameter to the [0,1] interval, with the following formula:

[0039] in, These are the original parameter values. and These are the minimum and maximum values ​​of the parameter in the training dataset, respectively. Normalization can eliminate the influence of differences in the units and magnitudes of different parameters on the clustering algorithm, improving the stability and classification accuracy of the fuzzy C-means (FCM) algorithm.

[0040] Specifically, this step is typically deployed on the monitoring platform of the fuel system in thermal power plants, achieving automated data collection and processing through the data interface between the DCS system and the host computer. The normalized data serves as input to the IPOA-FCM model, providing high-quality feature vectors for subsequent clustering training. The technical benefit of this step is that it provides a well-structured and standardized training dataset for the fault diagnosis model, thereby improving the model's ability to identify coal shortage faults and laying a data foundation for early warning and intelligent diagnosis.

[0041] Furthermore, S2 includes: S21, the collected data is segmented using a sliding window method, with a window length of [missing information]. Each sampling period, with an overlap rate of [number] times. .

[0042] Specifically, in this invention, the sliding window method described in step three for segmenting the collected data is a crucial step in realizing the structuring of the input data and the extraction of time-series features for the fault diagnosis model. This method, based on the continuity of time-series data and the correlation of local features, uses a fixed-length window to segment the collected operating parameters. The window length is... Each sampling period, with an overlap rate of [number] times. ,in By setting the window length It allows control over the temporal resolution and feature extraction granularity of each data segment, while the overlap rate... This is used to enhance the continuity between adjacent data segments and avoid the loss of feature information due to breakpoints between windows.

[0043] Specifically, the sliding window method first selects the collected raw data sequence Time alignment is performed according to the sampling period, followed by step size. A sliding window divides the data into several subsequences. ,in Each subsequence corresponds to a multidimensional operating parameter within a time window, such as feeder speed, current, and differential pressure at the inlet and outlet of the coal mill, providing structured input for subsequent cluster analysis.

[0044] Furthermore, window length The sampling period is typically set based on the system sampling frequency and the time-varying characteristics of the fault features. For example, in a thermal power plant, the sampling period is 1 second. Each data segment then covers 10 seconds of the running status. Overlap rate The value of affects the model's sensitivity to dynamic changes. It is generally set to 0.2~0.5 to ensure sufficient information overlap between adjacent windows and improve the model's ability to identify transition states.

[0045] Specifically, this step is widely applicable in real-time monitoring and fault diagnosis scenarios for fuel systems in thermal power units. Using the sliding window method, the system can transform continuously collected operating parameters into time-series data blocks suitable for cluster analysis, thus providing high-quality input data for the subsequent IPOA-FCM model. Its technical advantages lie in effectively extracting the local time-series features of operating parameters, enhancing the model's ability to identify coal shortage faults, and improving the continuity and stability of data processing, laying the foundation for achieving high-precision, low-false-rate fault diagnosis.

[0046] S22, performs wavelet denoising on the normalized data, preserving... The signal components.

[0047] Specifically, in the technical solution of this invention, step three involves performing wavelet denoising on the normalized data. The core objective of this step is to remove noise interference from the collected data and retain signal components with actual physical meaning, thereby improving the accuracy and robustness of subsequent clustering analysis. This step utilizes the multi-scale analysis characteristics of wavelet transform to perform joint time-frequency domain processing on the signal, achieving effective noise suppression.

[0048] Specifically, wavelet denoising typically involves three stages: signal decomposition, thresholding, and signal reconstruction. First, the normalized data is decomposed into multiple layers using Discrete Wavelet Transform (DWT), breaking the signal down into approximation coefficients and detail coefficients at different scales. The detail coefficients primarily reflect the high-frequency components of the signal and often contain noise information. Next, the detail coefficients are subjected to soft or hard thresholding to eliminate noise. The threshold can be based on a universal threshold or an adaptive thresholding method based on signal characteristics, such as Stein's Unbiased Risk Estimation (SURE). Finally, the processed coefficients are reconstructed into the denoised signal using Inverse Discrete Wavelet Transform (IDWT).

[0049] Specifically, this invention particularly emphasizes the retention of The signal components. Signal-to-noise ratio (SNR) is an important indicator of signal quality, defined as the ratio of signal power to noise power, calculated using the following formula: The unit is decibel (dB). In actual processing, by calculating the SNR values ​​of the signals before and after noise reduction, signal components with an SNR higher than 10dB are selected to ensure that the retained signals have sufficient clarity and representativeness, thereby providing high-quality input data for the subsequent IPOA-FCM clustering model.

[0050] Furthermore, this step is primarily applied to the real-time monitoring and fault diagnosis of the coal feeder's operating status in the fuel system of thermal power plants. Because the coal feeder is affected by various factors during operation, such as coal quality fluctuations, mechanical wear, and foreign object blockage, its operating parameters (such as current and speed) often contain a significant amount of noise. Wavelet denoising processing can effectively extract key features reflecting changes in equipment status, improving the accuracy of fault identification. This step is typically performed after data is collected from the DCS system and normalized, providing a stable and reliable signal input for subsequent cluster analysis.

[0051] Specifically, this step significantly improves the signal-to-noise ratio of the data and reduces the interference of noise on the clustering results, thereby improving the classification performance of the IPOA-FCM model under limited sample conditions. By retaining The signal components ensure the model's ability to identify states such as "normal," "transition," and "fault," providing a solid data foundation for early warning and intelligent diagnosis of coal feeder failures.

[0052] S3 employs an improved pelican optimization algorithm to optimize the initial centers of fuzzy C-means clustering, including initializing the number of clusters, IPOA parameters, and FCM parameters. The initial cluster centers are optimized by a population initialization strategy that simulates pelican predation behavior. The fitness function is defined using the FCM objective function, and particle iterative optimization is performed.

[0053] Specifically, in this invention, step three proposes to use an improved parrot optimization algorithm (IPOA) to optimize the initial centers of fuzzy C-means (FCM) clustering, thereby improving the convergence performance and classification accuracy of the clustering algorithm under limited sample conditions. The technical implementation of this step is based on a combination of swarm intelligence optimization and fuzzy clustering, specifically including key steps such as parameter initialization, population initialization strategy, fitness function definition, and particle iterative optimization.

[0054] Specifically, the number of clusters needs to be set first. Typically, the coal feeder's operating status is categorized into four types (normal, transitional, fault, and abnormal). IPOA parameters include population size. Maximum number of iterations Inertia weight Individual learning factors and group learning factors Their value ranges are respectively , , , The parameters of FCM include the fuzzy weighting index. Generally take and the error threshold for terminating the iteration. .

[0055] Furthermore, the population initialization strategy simulates pelican predation behavior, generating initial particle positions through a combination of random distribution and local perturbation to ensure good diversity in the solution space. The fitness function is defined using the objective function of FCM, which is to minimize the sum of squared errors in fuzzy clustering, and its expression is:

[0056] in, Indicates the first The nth sample pair Membership degree of each cluster center For sample vectors, As cluster center, For fuzzy index, Represents the Euclidean distance. IPOA optimizes the particle position by iteratively updating it. The initial value is used to improve the clustering performance of FCM.

[0057] Specifically, this method is applicable to intelligent identification of the operating status of coal feeders in thermal power plant fuel systems. Especially when sample data is limited and operating status changes are complex, it effectively avoids the problem of the FCM algorithm getting trapped in local optima due to improper initial center selection. Optimizing the initial centers through IPOA significantly improves the stability and accuracy of clustering results, providing reliable data support for subsequent fault diagnosis and alarm mechanisms. It has good engineering practical value and promising prospects for widespread application.

[0058] Furthermore, S3 includes: S31, generating initial cluster centers through a population initialization strategy simulating pelican predation behavior, specifically including: Specifically, this step generates initial cluster centers using a population initialization strategy that simulates pelican predation behavior. It is a crucial step in the clustering model construction process, which combines the Improved Parrot Optimization Algorithm (IPOA) with the Fuzzy C-Means (FCM) algorithm. In some implementations, this strategy effectively mitigates the FCM algorithm's sensitivity to initial cluster centers and its tendency to get trapped in local optima by constructing a diverse initial population, thereby improving the stability and accuracy of the clustering results.

[0059] Specifically, IPOA simulates the group behavior of pelicans during predation, including stages such as searching, encircling, and convergence. The algorithm first randomly initializes a certain number of "pelican" individuals in the normalized data space of the feature parameters, with each individual representing a potential cluster center. During iteration, individuals update their positions according to a fitness function (i.e., the objective function of FCM), defined as:

[0060] in, Indicates the first The nth sample pair Membership degree of each cluster center For fuzzy weighted index (usually taken as) ), For the input sample vector, The cluster center vector, The number of clusters, This represents the total number of samples. IPOA optimizes... The initial position of the objective function makes the objective function In subsequent FCM iterations, it quickly converges to the global optimum or suboptimal solution.

[0061] Furthermore, the population size of IPOA is typically set to... Number of iterations fitness threshold This is to ensure that the algorithm reaches a stable state within a reasonable time. Number of clusters The coal feeder's operating status is divided into three categories (normal, transition, and fault), and can be expanded to four categories to include abnormal data (marked as category "0").

[0062] Specifically, this step is mainly used in practical applications for fault diagnosis of fuel systems in thermal power plants, especially in the early identification of coal feeder failures. By collecting operating parameters from the DCS system (such as coal feeder current, coal mill inlet and outlet differential pressure, etc.) and performing normalization processing, the IPOA-FCM model can perform cluster analysis on the data to determine whether the coal feeder is in a normal, transitional, or faulty state.

[0063] Specifically, this population initialization strategy significantly improves the global search capability of the clustering algorithm and reduces the clustering bias caused by improper initial center selection in the traditional FCM algorithm. Simultaneously, by simulating the group behavior of pelicans, IPOA maintains population diversity during optimization, avoiding premature convergence and thus improving the model's accuracy and robustness in identifying coal shortage faults. This step provides high-quality clustering results for subsequent fault diagnosis and alarm mechanisms, forming the core foundation for intelligent fault early warning.

[0064] S32, with FCM objective function Define the fitness function, where For fuzzy membership, As cluster center, It is a fuzzy weighted index.

[0065] Specifically, in the technical solution of this invention, step three defines a fuzzy C-means (FCM) objective function. The fitness function, used in the Improved Parrot Optimization Algorithm (IPOA), is employed to optimize the selection of initial cluster centers. At the technical implementation level, this step first requires clarifying the mathematical structure of the FCM objective function and its role in the clustering process. Objective Function Characterizes the data points With cluster center The sum of fuzzy weighted distances between them, where Indicates the first The data point pair Fuzzy membership degree of a cluster, It is a fuzzy weighted index, and its value range is typically [value range missing]. In this invention, it is preferred that... This is to enhance the impact of fuzziness on the clustering results. In the IPOA optimization process, the definition of the fitness function directly determines the algorithm's search direction for cluster centers and its optimization efficiency.

[0066] Specifically, the number of clusters in the FCM algorithm Typically, the classification is set according to the actual operational status requirements; in this invention, it is set as follows: These correspond to "normal coal feeder operation", "coal shortage transition state", "coal shortage fault state", and "abnormal unclassified state", respectively. Initial cluster centers Optimization using IPOA yields a fitness function that is the objective function described above. By minimizing To find the optimal distribution of cluster centers. The population size, number of iterations, step size, and other parameters of IPOA need to be adjusted according to the actual dataset size and computing resources. Typically, the population size is set to [value missing]. The maximum number of iterations is This is to ensure that the algorithm converges within a reasonable time.

[0067] Furthermore, this step is primarily used in intelligent diagnostic systems for coal feeder failures. By inputting collected operating parameters (such as feeder current, speed, and mill inlet / outlet differential pressure) into the IPOA-FCM model, automatic classification of operating states is achieved. In actual industrial environments, this method can be deployed in DCS systems, combining real-time data streams for online cluster analysis, thereby improving the real-time performance and accuracy of fault identification.

[0068] Specifically, this step effectively guides IPOA to search for the optimal initial cluster centers in the multidimensional parameter space by using the FCM objective function as the fitness function. This avoids the sensitivity of the traditional FCM algorithm to initial values ​​and improves clustering stability and diagnostic accuracy. Furthermore, this method maintains good classification performance even with limited sample data, providing reliable technical support for early warning of coal feeder failures.

[0069] S4. Based on the optimized initial cluster centers, perform FCM clustering iteration. Determine the operating status of the coal feeder according to the Euclidean distance between the monitored feature parameters and the standard cluster center points. Divide the status into four categories: normal (1), coal interruption transition (2), coal interruption fault (3), and abnormal (0).

[0070] Specifically, in this invention, performing FCM clustering iteration based on the optimized initial cluster centers is one of the core steps in realizing intelligent diagnosis of the coal feeder's operating status. This step is performed after the IPOA optimization algorithm completes the global search for the initial cluster centers, aiming to achieve accurate identification of the coal feeder's operating status by dynamically clustering the monitored feature parameters using the Fuzzy C-Means (FCM) algorithm.

[0071] Specifically, the FCM algorithm minimizes the objective function. To iteratively optimize the cluster center and membership matrix, the objective function is defined as:

[0072] in, For the sample size, The number of clusters (4 in this invention). For fuzzy weighted index (usually taken as) ), Indicates the first The nth sample pair Membership degree of each cluster center For the first The feature vector of each sample For the first The initial cluster centers. It is obtained by optimization of the improved Pelican Algorithm (IPOA), which effectively avoids the problem of traditional FCM algorithm getting stuck in local optima due to improper initial center selection.

[0073] Specifically, in this invention, the iteration termination condition of the FCM algorithm is typically set to the change in the membership matrix being less than a preset threshold. (like (or reaching the maximum number of iterations) (like Cluster number The four operating states of the coal feeder are: normal (1), coal interruption transition (2), coal interruption fault (3), and abnormal (0). The state classification is based on the Euclidean distance between the monitored characteristic parameters and the center point of the standard class. ,when Exceeding the set threshold When this occurs, it is determined to be an "abnormal" state.

[0074] Furthermore, this step is deployed in the real-time monitoring and diagnostic module of the power plant's fuel system. The monitoring system collects characteristic parameters including feeder speed, current, pulverizer current, and primary air flow, which are then normalized and input into the IPOA-FCM model. The diagnostic system uses clustering results to determine in real time whether the feeder is in a coal supply interruption transition or coal supply interruption fault state, and triggers an audible and visual alarm module when state "2" or "3" is identified, achieving early warning and intervention for faults.

[0075] Specifically, by optimizing the initial cluster centers, the clustering stability and accuracy of the FCM algorithm under small sample conditions are significantly improved, thereby enhancing the reliability of fault diagnosis. Meanwhile, the state partitioning mechanism based on Euclidean distance gives the system strong robustness and real-time response capabilities, providing strong support for the intelligent operation and maintenance of power plant fuel systems.

[0076] Furthermore, S4 includes: S41, Calculate monitoring characteristic parameters Center point of standard class European distance ,in , .

[0077] Specifically, in the steps of this invention, monitoring characteristic parameters are calculated. Center point of standard class European distance This step is the core component of the cluster analysis for the coal feeder's coal shortage fault diagnosis model. Based on the fuzzy C-means (FCM) algorithm, it measures the similarity between each sample point and the cluster center, thus providing a basis for subsequent fault state classification.

[0078] Furthermore, the calculation of Euclidean distance uses the standard multidimensional spatial distance formula, namely... ,in Indicates the first One monitoring sample point, Indicates the first A standard class center point The dimension of the feature parameters is denoted as . In this invention, the feature parameters include feeder speed, feeder current, mill current, primary air flow rate, inlet and outlet differential pressure, outlet temperature, outlet air-coal pressure, and inlet primary air pressure, totaling eight dimensions. By calculating the Euclidean distance between each sample point and the center point of each class, its fuzzy membership degree to a certain class can be evaluated, thereby achieving the classification of the feeder's operating status.

[0079] Specifically, the accuracy of Euclidean distance calculation directly affects the stability of clustering results and the accuracy of classification. In this invention, all feature parameters are normalized before being input into the model to eliminate the influence of dimensional differences on distance calculation. The normalization method can be either Min-Max normalization or Z-Score normalization, ensuring the data distribution is even. Or within a range where the mean is 0 and the standard deviation is 1. Cluster centers The initial value is obtained by optimization using the improved Pelican Algorithm (IPOA) to improve clustering convergence speed and avoid local optima.

[0080] Furthermore, this step is embedded in the real-time monitoring system for status identification of the collected operational data. When the monitoring system acquires the characteristic parameters of the current moment... Then, the diagnostic system will calculate its relationship with various center points. The model is evaluated by calculating the Euclidean distance between the two values ​​and determining their category based on the minimum distance principle or a fuzzy membership function. If the distance exceeds a preset threshold, it is considered an abnormal state (category "0"), and the model update mechanism is triggered.

[0081] Specifically, precise distance calculation provides a reliable similarity metric for fuzzy clustering, thereby enabling high-precision identification of the coal feeder's operating status. Combining this with an improved IPOA algorithm to optimize the initial cluster centers further enhances the model's robustness and classification accuracy under limited sample conditions, providing a solid foundation for early warning and status monitoring of coal supply failures.

[0082] S42, according to the preset threshold Classify, when When the condition is determined to be abnormal (0), it is considered an abnormal state.

[0083] Specifically, this step is a key classification and judgment step in the data-driven coal feeder coal shortage fault diagnosis method. Its technical implementation principle is based on the model output results of a combination of fuzzy C-means (FCM) clustering algorithm and an improved Pelican Optimization Algorithm (IPOA). After completing the clustering training, the system will use the real-time collected coal feeder operating characteristic parameters... With cluster center Perform Euclidean distance calculations to determine whether the current operating status is abnormal.

[0084] The specific operation method is as follows: First, the system obtains the feature parameter vector at the current moment through the monitoring module. This vector contains key variables such as feeder speed, current, mill current, and primary air flow. Subsequently, With cluster center Perform Euclidean distance calculation, i.e. If the distance is greater than a preset threshold If the threshold is not met, the sample point is determined to belong to class "0", i.e., an abnormal state. Typically, the settings are based on the statistical distribution of historical data, for example, by using boundary analysis of clustering results during the training phase. Set as ,in This represents the standard deviation between the normal class and the transition class.

[0085] Furthermore, the monitoring system collects data through the DCS engineering station, while the diagnostic system performs status identification based on the IPOA-FCM model. When the operating parameters of a coal feeder deviate from the normal range and the Euclidean distance exceeds... When this happens, the system will trigger an alarm mechanism to prompt operators to check or intervene.

[0086] Specifically, by introducing a threshold determination mechanism, normal, transitional, fault, and abnormal states are effectively distinguished, improving the robustness and accuracy of fault identification. Especially when sample data is limited, it avoids the dependence of traditional methods on a large number of fault samples, while enhancing the ability to identify unknown abnormal patterns.

[0087] S5, when the diagnosis result is a coal shortage transition state or a coal shortage fault state, the audible and visual alarm module is triggered to generate alarm information. If the Euclidean distance between the monitored characteristic parameter and the standard class center point exceeds the preset threshold, it is classified as an abnormal state and the model parameters are updated.

[0088] Specifically, when the diagnostic system identifies that the coal feeder is in a coal shortage transition state (state "2") or a coal shortage fault state (state "3"), the system will trigger the audible and visual alarm module to generate corresponding alarm information. This step determines the state based on the output of the clustering diagnostic model and combines it with the Euclidean distance threshold determination mechanism to achieve the identification of abnormal states and the dynamic updating of model parameters.

[0089] Furthermore, the audible and visual alarm module is controlled collaboratively by hardware and software. The hardware includes a high-decibel buzzer and a high-brightness LED warning light, while the software uses state recognition logic and a threshold judgment algorithm for control. When the diagnostic system outputs a state of "2" or "3", the system activates the alarm module, causing it to simultaneously emit a sharp alarm sound and a flashing warning light to ensure that operators can respond quickly and take appropriate measures. The alarm signal triggering mechanism is based on a dual judgment of state classification results and Euclidean distance, where the Euclidean distance is calculated using the following formula:

[0090] in, This represents the first monitored characteristic parameter. Each dimension value The first point representing the center of the standard class Each dimension value This represents the total dimension of the feature parameters. If... Exceeding the preset threshold If the system detects an abnormal state "0", it will update the model parameters according to the actual operating conditions to adapt to the dynamic changes in the equipment status.

[0091] Specifically, the Euclidean distance threshold Typically, the settings are based on the statistical distribution of historical data, such as using the three-standard-deviation method or inter-class distance analysis based on clustering results. The response time of the audible and visual alarm module should be controlled within 50ms to meet the real-time requirements of industrial sites. The brightness of the LED light should not be less than 1000mcd, and the output frequency of the buzzer is recommended to be set between 2000Hz and 4000Hz to enhance the human ear's perception.

[0092] Specifically, this step is widely applicable to the monitoring of the operating status of coal feeders and coal mills in the fuel systems of thermal power plants. Especially under conditions where coal shortage accidents occur frequently, the system can achieve early warning of transition from a transitional state to a fault state through real-time monitoring and intelligent diagnosis, thereby reducing equipment downtime and maintenance costs.

[0093] Specifically, by combining audible and visual alarm mechanisms with Euclidean distance determination, the accuracy of fault identification is improved, and the system's real-time response and adaptive capabilities are enhanced. With the support of a model parameter update mechanism, the system can continuously optimize diagnostic performance and adapt to the evolution of equipment operating conditions, thereby achieving more efficient and intelligent fault diagnosis and early warning functions.

[0094] Furthermore, S5 includes: S51 uses a combination of LED lights and a buzzer to achieve an audible and visual alarm. The LED lights use red and green dual-color indicator lights. A solid red light indicates a coal shortage fault state (3), and a flashing red light indicates a coal shortage transition state (2).

[0095] Specifically, this step utilizes a combination of LED lights and a buzzer to achieve an audible and visual alarm function. This is a key output module in the data-driven coal feeder coal shortage fault diagnosis method and alarm system of this invention. Technically, the audible and visual alarm module consists of a red-green dual-color LED light group and a high-pitched buzzer. The LED light group adopts a common anode or common cathode structure and is controlled by a microcontroller (such as an ARM Cortex-M series or PLC). A solid red light indicates that the coal feeder is in a coal shortage fault state (state 3); a periodic flashing red light at a frequency of 1Hz indicates a transitional coal shortage state (state 2); and a solid green light indicates that the system is operating normally (state 1). Upon receiving a fault or transitional state signal, the buzzer emits a sharp alarm sound at a frequency of 2kHz for 10 seconds to ensure that operators can quickly detect abnormalities in noisy industrial environments.

[0096] Specifically, the driving current of the LED light assembly is controlled within 20mA to meet the power consumption and lifespan requirements of industrial equipment; the sound pressure level of the buzzer is not less than 85dB(A) at a distance of 1 meter, which meets the acoustic performance requirements for emergency alarm devices in GB 4798.7-1984 "Industrial Environmental Noise Control Standard". The alarm triggering logic is based on the output state of the IPOA-FCM clustering model. When the diagnostic system identifies a state of "2" or "3", it sends a control signal to the alarm module through the GPIO interface to achieve synchronous sound and light alarm.

[0097] Furthermore, this alarm system is deployed in the fuel delivery system of thermal power plants and integrated with the DCS system to monitor the operating status of the coal feeder in real time. When the system detects a coal shortage fault or transitional state, the alarm module responds immediately, providing operators with intuitive visual and auditory prompts, thereby enabling rapid intervention and preventing the coal mill from shutting down or being damaged due to coal shortage.

[0098] Specifically, the dual audible and visual alarm mechanism significantly improves the timeliness and identifiability of fault response, enhances the system's reliability and human-computer interaction efficiency, and provides an intuitive and effective feedback means for achieving intelligent and automated fault diagnosis.

[0099] S52, according to Calculate signal transmission delay to ensure real-time alarm information, where For transmission power, For transmission distance, For bandwidth, This refers to the signal-to-noise ratio.

[0100] Specifically, in the technical solution of the present invention, step four involves applying the formula... Calculate the signal transmission delay to ensure the real-time nature of alarm information. This step plays a crucial role in data-driven coal feeder coal shortage fault diagnosis systems, especially in time control between fault identification and alarm response.

[0101] Specifically, this formula is used to quantify the transmission delay between the detection of a coal shortage fault signal by the diagnostic system and the delivery of the alarm information to the control terminal. .in, This indicates the signal transmission power, measured in watts (W). This represents the signal transmission distance, measured in meters (m). This refers to communication bandwidth, measured in Hertz (Hz). Signal-to-noise ratio (SNR) is usually expressed in decibels (dB), but it needs to be converted to a linear ratio form in the formula. This formula combines key factors such as signal transmission capability, transmission distance, communication bandwidth, and channel quality, and can predict the transmission time of alarm signals relatively accurately under specific communication conditions.

[0102] Furthermore, Typically set in the range of 10mW to 1W to suit wireless communication modules in industrial settings; Depending on the layout of the equipment on site, it generally does not exceed 100m; Typically, it ranges from 1MHz to 10MHz, depending on the wireless communication protocol used (such as ZigBee, LoRa, or 5G industrial IoT). In industrial environments, the signal strength must be maintained above 10 dB to ensure reliable signal transmission. By adjusting these parameters, the system can optimize communication link performance, thereby reducing... Improve alarm response speed.

[0103] Specifically, this step is mainly used in the fuel system of thermal power plants. When the coal feeder is in a coal shortage transition state or a fault state, the diagnostic system needs to transmit alarm information to the control center or on-site operators in the shortest possible time. This is achieved through real-time calculations. The system can dynamically assess the latency of the current communication link and trigger a link optimization mechanism or switch the communication mode when the latency exceeds a preset threshold, so as to ensure that alarm information is delivered in a timely manner.

[0104] Specifically, by quantifying signal transmission delay, the system can effectively ensure the real-time nature of alarm information, thereby improving fault response efficiency. In actual operation, if... Exceeding the set response time limit (e.g., 50ms) may lead to malfunctions or delays, affecting equipment safety and production efficiency. Therefore, this formula provides a crucial performance evaluation basis for the system, enhancing the reliability and practicality of the overall diagnostic and alarm system.

[0105] S6, dynamically adjusts the Euclidean distance threshold based on historical operating data. Specifically, it includes: Specifically, this step involves dynamically adjusting the Euclidean distance threshold based on historical operational data. Its core lies in improving the adaptability and robustness of fault diagnosis models through data-driven approaches. In some implementations, this threshold is used to measure the deviation between the current monitored feature parameters and the standard cluster centroids, thereby determining whether the coal feeder is in a normal, transitional, or faulty state. Specifically, when the Euclidean distance between the feature vector and all cluster centroids at a certain moment is greater than... When this happens, the system classifies the state as "0", indicating an abnormal or unknown state, requiring further model updates or manual intervention.

[0106] Specifically, this step involves threshold adjustment based on the clustering results of the Fuzzy C-means (FCM) algorithm. First, multiple feature parameters (such as feeder speed, current, and mill inlet / outlet differential pressure) are extracted from historical operating data and divided into three categories: normal, transitional, and faulty samples. After model training, the system calculates the cluster centroid for each category and statistically analyzes the Euclidean distance distribution between each category of samples and its centroid. By setting a sliding window mechanism, the system can periodically update the statistical characteristics of the distance distribution (such as mean, variance, and maximum value) and dynamically adjust accordingly. The value of . Optionally, Can be set to distance distribution Quantiles are used to ensure that the vast majority of normal samples are not misclassified as abnormal.

[0107] Furthermore, The adjustment frequency can be set from once every 10 minutes to once per hour, depending on the system response requirements, specifically the real-time data acquisition and system stability requirements. In practical applications, this step is typically deployed in the power plant's DCS system, combined with real-time monitoring and diagnostic modules, to achieve continuous assessment and early warning of the coal feeder's operating status.

[0108] Specifically, this step effectively solves the problem of poor adaptability of the traditional fixed threshold method under different operating conditions, improving the accuracy and timeliness of fault identification. Through dynamic adjustment... The system can adapt to the characteristics of operating parameters changing over time, thus maintaining high diagnostic sensitivity and specificity under complex operating conditions, providing a reliable basis for subsequent audible and visual alarms and maintenance decisions.

[0109] This invention provides a method for diagnosing coal feeder failures based on an improved pelican algorithm, which dynamically adjusts the Euclidean distance threshold according to historical operating data. This further enhances the adaptability and robustness of the improved Pelican algorithm in diagnosing coal feeder failures, effectively responding to changes in operating conditions and improving the sensitivity of early identification and the reliability of real-time alarms.

[0110] Example 2 This invention also provides an alarm system for diagnosing coal feeder failures based on an improved pelican algorithm, such as... Figure 2 As shown, the implementation of this system includes the following steps: S101: Select model input parameters. The model input parameters, after analysis and screening, include: coal feeder speed, coal feeder current, coal mill current, coal mill primary air flow rate, coal mill inlet and outlet differential pressure, coal mill outlet temperature, coal mill outlet air-coal pressure, and coal mill inlet primary air pressure.

[0111] S102: Data Acquisition and Preprocessing. Data on various characteristic parameters during the coal shortage incident were collected through the DCS engineering station, and the parameters were divided into three different stages: normal, transition, and fault, and then normalized.

[0112] S103: Establish an IPOA-FCM coal feeder coal shortage fault diagnosis model. The construction of the IPOA-optimized FCM model includes: 1. Initializing parameters, setting the number of clusters, IPOA parameters, and FCM parameters; 2. Optimizing the initial cluster centers with IPOA, including population initialization, defining the fitness function based on the FCM objective function, and particle iterative optimization; 3. FCM clustering iteration, executing the FCM algorithm using the initial cluster centers obtained from IPOA optimization.

[0113] S104: Establish a coal feeder coal shortage fault alarm system. The alarm system is divided into a monitoring system and a fault diagnosis system. The monitoring system displays and stores the operating data of characteristic parameters in real time. The diagnosis system obtains the operating data of characteristic parameters from the monitoring system to judge the operating status of the coal feeder. When the diagnosis system detects that the coal feeder is in a coal shortage transition state, it triggers the audible and visual alarm module to generate alarm information.

[0114] Furthermore, the current operating status of the coal feeder is determined based on data-driven analysis, such as... Figure 3 As shown, it includes the following steps: S111: Collect relevant datasets of coal shortage accidents at power plants, and perform data segmentation and normalization.

[0115] S112: Substitute the dataset into the IPOA-FCM model for cluster training.

[0116] S113: The coal feeder coal shortage fault diagnosis model, trained on the dataset, is used to determine the coal feeder's operating status. The coal feeder's operating status is divided into four categories: "1" indicates "coal feeder is operating normally"; "2" indicates "coal feeder is in a coal shortage transition state"; "3" indicates "coal feeder is in a coal shortage fault state"; and "0" indicates that the Euclidean distance between the monitored feature parameters and the standard class center point exceeds the set threshold.

[0117] S114: The coal feeder coal shortage fault diagnosis model outputs the diagnosis results. If category "2" or "3" occurs, an audible and visual alarm is triggered. If category "0" occurs, the model status is updated according to the actual situation.

[0118] Furthermore, the coal feeder coal shortage fault diagnosis model was selected based on IPOA-FCM clustering training. IPOA can establish a diverse population from the beginning, reducing the risk of getting trapped in local optima, dynamically adapting to the needs of global exploration and local development, improving the ability to handle complex fault problems, while maintaining the algorithm's efficiency and robustness. FCM is a commonly used clustering analysis algorithm. As an unsupervised learning method, it has low dependence on training samples, fast computation speed, and is more suitable for engineering applications.

[0119] Furthermore, such as Figure 4 As shown, an alarm system for diagnosing coal feeder failure based on an improved pelican algorithm also includes the following subsystems: Specifically, this includes monitoring systems, diagnostic systems, and audible and visual alarms.

[0120] Preferably, the monitoring system displays and stores the operational data of characteristic parameters in real time. The diagnostic system acquires the operational data of characteristic parameters from the monitoring system to determine the operating status of the coal feeder. When the diagnostic system detects that the coal feeder is in a coal-cut-off transition state or a coal-cut-off state, it triggers the audible and visual alarm module to generate alarm information.

[0121] Preferably, the alarm is an audible and visual alarm that can emit a sharp sound and light simultaneously when triggered.

[0122] This invention discloses an alarm system for diagnosing coal feeder failures based on an improved pelican algorithm. By constructing a multi-dimensional dynamic feature space of operating parameters and employing the improved pelican optimization algorithm to optimize the initial centers of fuzzy C-means clustering, it effectively overcomes the core shortcomings of traditional methods, such as being prone to getting trapped in local optima and being sensitive to initial values. This method achieves full automation from data acquisition, feature extraction, cluster optimization to state recognition and early warning, significantly improving the accuracy and real-time performance of fault diagnosis. By establishing a complete monitoring-diagnosis-alarm closed-loop system, it can accurately identify the normal operation, coal feeder transition, and fault states of the coal feeder in complex industrial scenarios and promptly trigger graded alarms, thereby greatly improving equipment reliability, reducing maintenance costs, and enhancing the system's adaptability and practical value in engineering practice.

[0123] Example 3 This invention also provides a coal feeder coal shortage fault diagnosis device 10 based on an improved pelican algorithm, such as... Figure 5 As shown, the device includes: The multi-dimensional parameter input module 100 is used to select multi-dimensional operating parameters, including coal feeder speed, coal feeder current, coal mill current, coal mill primary air flow rate, coal mill inlet and outlet differential pressure, coal mill outlet temperature, coal mill outlet air-coal pressure and coal mill inlet primary air pressure, as model input features to construct a dynamic feature space.

[0124] The data acquisition and preprocessing module 200 is used to collect operational data during coal shortage accidents through the DCS system, divide the data into three stages: normal, transition, and fault, and perform normalization processing.

[0125] Specifically, the collected data is segmented using a sliding window method, with a window length of [missing information]. Each sampling period, with an overlap rate of [number] times. Perform wavelet denoising on the normalized data, retaining... The signal components.

[0126] The cluster center optimization module 300 is used to optimize the initial centers of fuzzy C-means clustering using an improved pelican optimization algorithm. This includes initializing the number of clusters, IPOA parameters, and FCM parameters. The initial cluster centers are optimized using a population initialization strategy simulating pelican predation behavior, with the FCM objective function as the target function. Define the fitness function and perform particle iterative optimization.

[0127] Specifically, initial cluster centers are generated using a population initialization strategy that simulates pelican predation behavior, including: randomly initializing the population position vector. ,in , For population size; according to Update individual location, where This is the step size coefficient. This is the current optimal solution; using the FCM objective function Define the fitness function and perform particle iterative optimization, where For fuzzy membership, As cluster center, It is a fuzzy weighted index.

[0128] The state classification module 400 is used to perform FCM clustering iterations based on the optimized initial cluster centers, according to the Euclidean distance between the monitored feature parameters and the standard cluster centers. The operating status of the coal feeder is determined and classified into four categories: normal (1), coal interruption transition (2), coal interruption fault (3), and abnormal (0).

[0129] Specifically, calculate monitoring characteristic parameters Center point of standard class European distance ,in , According to the preset threshold Classify, when When the condition is determined to be abnormal (0), it is considered an abnormal state.

[0130] The alarm and parameter update module 500 is used to trigger the audible and visual alarm module to generate alarm information when the diagnostic result indicates a coal shortage transition state or a coal shortage fault state. If the Euclidean distance between the monitored characteristic parameter and the standard class center point exceeds a preset threshold... If it does, it is classified as an abnormal state and the model parameters are updated.

[0131] Specifically, an audible and visual alarm is achieved by combining LED lights with a buzzer. The LED lights use red and green dual-color indicator lights. A solid red light indicates a coal shortage fault state (3), and a flashing red light indicates a coal shortage transition state (2). Calculate signal transmission delay to ensure real-time alarm information, where For transmission power, For transmission distance, For bandwidth, This refers to the signal-to-noise ratio.

[0132] Specifically, a coal feeder coal shortage fault diagnosis device based on the improved pelican algorithm further includes: a dynamic threshold adjustment module, used to dynamically adjust the Euclidean distance threshold according to historical operating data. Specifically, this includes: calculating the mean Euclidean distance of normal state samples within the current period. and standard deviation ;according to Update threshold, where , .

[0133] This invention discloses a coal feeder coal shortage fault diagnosis device based on an improved pelican algorithm. By constructing a multi-dimensional dynamic feature space of operating parameters and employing the improved pelican optimization algorithm to optimize the initial cluster centers of fuzzy C-means clustering, it effectively overcomes the core defects of traditional methods, such as sensitivity to initial values ​​and susceptibility to local optima. The device achieves full automation from data acquisition and processing, cluster center optimization, operating status identification to early warning feedback, significantly improving the accuracy and real-time performance of fault diagnosis. By establishing a complete monitoring-diagnosis-alarm closed-loop system, it can accurately identify different operating states of the coal feeder in complex industrial environments and promptly trigger tiered alarms. It also possesses parameter self-updating capabilities, thereby greatly improving equipment reliability and system adaptability, and effectively reducing operation and maintenance costs.

[0134] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0135] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present 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.

[0136] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

Claims

1. A method for diagnosing coal feeder failure based on an improved pelican algorithm, characterized in that, include: S1. Select multi-dimensional operating parameters including coal feeder speed, coal feeder current, coal mill current, coal mill primary air flow, coal mill inlet and outlet differential pressure, coal mill outlet temperature, coal mill outlet air-coal pressure and coal mill inlet primary air pressure as model input features to construct a dynamic feature space. S2 collects operational data during coal shortage accidents through the DCS system, divides the data into three stages: normal, transition, and fault, and performs normalization processing. S3 uses an improved pelican optimization algorithm to optimize the initial centers of fuzzy C-means clustering, including initializing the number of clusters, IPOA parameters, and FCM parameters. The initial cluster centers are optimized by a population initialization strategy that simulates pelican predation behavior. The fitness function is defined by the FCM objective function and particle iterative optimization is performed. S4. Based on the optimized initial cluster centers, perform FCM clustering iteration. Determine the operating status of the coal feeder according to the Euclidean distance between the monitored feature parameters and the standard cluster center points. Divide the status into four categories: normal (1), coal interruption transition (2), coal interruption fault (3), and abnormal (0). S5, when the diagnosis result is a coal shortage transition state or a coal shortage fault state, the audible and visual alarm module is triggered to generate alarm information. If the Euclidean distance between the monitored characteristic parameter and the standard class center point exceeds the preset threshold, it is classified as an abnormal state and the model parameters are updated.

2. The method as described in claim 1, characterized in that, The improved Pelican optimization algorithm for optimizing the initial centers of fuzzy C-means clustering also includes: S31, generating initial cluster centers through a population initialization strategy simulating pelican predation behavior, specifically including: Randomly initialize population position vector ,in , Population size; according to Update individual location, where This is the step size coefficient. This is the current optimal solution; S32, with FCM objective function Define the fitness function, where For fuzzy membership, As cluster center, It is a fuzzy weighted index.

3. The method as described in claim 1, characterized in that, Performing FCM clustering iterations based on the optimized initial cluster centers also includes: S41, Calculate monitoring characteristic parameters Center point of standard class European distance ,in , ; S42, according to the preset threshold Classify, when When the condition is determined to be abnormal (0), it is considered to be in an abnormal state.

4. The method as described in claim 1, characterized in that, The audible and visual alarm module that triggers the alarm when the diagnosis result is a coal shortage transition state or a coal shortage fault state also includes: S51, an audible and visual alarm is achieved by combining LED lights with a buzzer. The LED lights use red and green dual-color indicator lights. A solid red light indicates a coal shortage fault state (3), and a flashing red light indicates a coal shortage transition state (2). S52, according to Calculate signal transmission delay to ensure real-time alarm information, where For transmission power, For transmission distance, For bandwidth, This refers to the signal-to-noise ratio.

5. The method as described in claim 1, characterized in that, The operational data collected during coal shortage accidents through the DCS system also includes: S21, the collected data is segmented using a sliding window method, with a window length of [missing information]. Each sampling period, with an overlap rate of [number] times. ; S22, performs wavelet denoising on the normalized data, preserving... The signal components.

6. The method as described in claim 1, characterized in that, Also includes: S6, dynamically adjusts the Euclidean distance threshold based on historical operating data. Specifically, it includes: Calculate the mean Euclidean distance of normal state samples within the current period. and standard deviation ; according to Update threshold, where , .

7. A coal feeder coal shortage fault diagnosis device based on an improved pelican algorithm, characterized in that, include: The multi-dimensional parameter input module is used to select multi-dimensional operating parameters, including coal feeder speed, coal feeder current, coal mill current, coal mill primary air flow rate, coal mill inlet and outlet differential pressure, coal mill outlet temperature, coal mill outlet air-coal pressure and coal mill inlet primary air pressure, as model input features to construct a dynamic feature space. The data acquisition and preprocessing module is used to collect operational data during coal shortage accidents through the DCS system, divide the data into three stages: normal, transition, and fault, and perform normalization processing. The cluster center optimization module is used to optimize the initial centers of fuzzy C-means clustering using an improved pelican optimization algorithm. This includes initializing the number of clusters, IPOA parameters, and FCM parameters. The initial cluster centers are optimized using a population initialization strategy simulating pelican predation behavior, with the FCM objective function as the target function. Define the fitness function and perform particle iterative optimization; The state classification module is used to perform FCM clustering iterations based on the optimized initial cluster centers, according to the Euclidean distance between the monitored feature parameters and the standard cluster centers. The operating status of the coal feeder is determined and classified into four categories: normal (1), coal interruption transition (2), coal interruption fault (3), and abnormal (0). The alarm and parameter update module is used to trigger the audible and visual alarm module to generate alarm information when the diagnostic result indicates a coal shortage transition state or a coal shortage fault state. This is done if the Euclidean distance between the monitored characteristic parameter and the standard class center point exceeds a preset threshold. If it does, it is classified as an abnormal state and the model parameters are updated.

8. The apparatus as claimed in claim 7, characterized in that, The cluster center optimization module is also used for: Initial cluster centers are generated using a population initialization strategy that simulates pelican predation behavior, specifically including: Randomly initialize population position vector ,in , Population size; according to Update individual location, where This is the step size coefficient. This is the current optimal solution; Using FCM objective function Define the fitness function and perform particle iterative optimization, where For fuzzy membership, As cluster center, It is a fuzzy weighted index.

9. The apparatus as claimed in claim 7, characterized in that, The state classification module is also used for: Calculate monitoring characteristic parameters Center point of standard class European distance ,in , ; According to the preset threshold Classify, when When the condition is determined to be abnormal (0), it is considered to be in an abnormal state.

10. The apparatus as claimed in claim 7, characterized in that, The alarm and parameter update module is also used for: The sound and light alarm is achieved by combining LED lights with a buzzer. The LED lights use red and green dual-color indicator lights. A solid red light indicates a coal shortage fault state (3), and a flashing red light indicates a coal shortage transition state (2). according to Calculate signal transmission delay to ensure real-time alarm information, where For transmission power, For transmission distance, For bandwidth, This refers to the signal-to-noise ratio.

Images