Method for realizing boiler wall temperature anomaly detection and early warning based on sample entropy calculation

Abstract

The invention discloses a method for achieving boiler wall temperature anomaly detection and early warning based on sample entropy calculation, and belongs to the technical field of boiler operation and maintenance. Comprising the following steps: step 1, dividing a sliding window, and observing temperature time sequence data by using the sliding window; step 2, calculating statistics in each sliding window; 3, constructing a parameter matrix according to the statistics; and 4, according to the parameter matrix, performing anomaly detection and early warning by using an OCSVM algorithm. According to the method, the detection precision and robustness of the abnormal data of the boiler wall temperature are improved, and various types of abnormities of the boiler wall temperature can be effectively detected.

Description

Technical Field

[0001] The present invention relates to the technical field of boiler operation and maintenance, and in particular to a method for detecting and warning abnormal boiler wall temperature based on sample entropy calculation. Background Art

[0002] Abnormal boiler wall temperatures in large coal-fired power plants can have serious safety implications in multiple ways. Excessively high boiler wall temperatures can alter the metallographic structure of metal materials, leading to phenomena such as pearlite spheroidization and graphitization, resulting in a decrease in material strength and hardness. When the material's strength is lower than the pressure and stress it withstands, deformation and cracking can occur. Abnormal wall temperatures can also reduce the toughness of metal materials, making them more brittle and weakening their impact resistance. During operation, exposure to external forces such as pressure fluctuations and thermal shock can easily lead to brittle fractures, potentially causing safety hazards. Abnormally high wall temperatures can alter the heat exchange conditions between the high-temperature flue gas in the combustion chamber and the heating surfaces, resulting in incomplete fuel combustion and making the resulting ash more likely to adhere to the heating surfaces, forming coke and ash deposits. This not only reduces the heat transfer efficiency of the heating surfaces but also further increases wall temperatures, creating a vicious cycle. Abnormal wall temperatures can also cause changes in the surface hardness of the metal materials on the heating surfaces, accelerating wear due to the erosion of fly ash particles in the flue gas. This can lead to thinning of the heating surface tube walls, reducing load-bearing capacity and posing safety hazards such as leaks.

[0003] Monitoring the boiler wall temperature in large coal-fired power plants is extremely important. Large coal-fired power plant boilers operate at high temperatures and high pressures. Wall temperature monitoring provides real-time visibility into the temperature of the boiler's heating surface. Any abnormal increase or decrease in wall temperature can promptly identify potential safety hazards, such as material degradation, creep, cracks, and even tube bursts caused by overheating. This prevents equipment damage and accidents, ensuring safe and stable operation of the power plant. A reasonable wall temperature distribution helps optimize the boiler's combustion and heat transfer processes. By monitoring wall temperature, operators can adjust combustion conditions based on temperature data to ensure full fuel combustion, improve thermal efficiency, and reduce energy waste. Furthermore, reduced heat transfer efficiency caused by factors such as slagging and ash accumulation can be promptly identified, allowing appropriate measures to be taken to maintain efficient boiler operation. Prolonged abnormal wall temperatures can accelerate aging and damage to the boiler's heating surface materials. Accurately monitoring and controlling wall temperature within a reasonable range can prevent fatigue damage to materials caused by overheating or undercooling, thereby extending the service life of boiler equipment and reducing equipment replacement and maintenance costs. Wall temperature monitoring data serves as a valuable basis for fault diagnosis. When a boiler malfunctions, analyzing the wall temperature's changing trends, distribution, and relationship with other operating parameters can help technicians quickly and accurately determine the fault's type, location, and cause. This allows for effective repair measures, shortening troubleshooting time and improving the plant's overall operational reliability. Extensive wall temperature monitoring data can provide a wealth of information for power plant operations management. By analyzing and mining this data, it's possible to determine the patterns of boiler wall temperature variation under different operating conditions, providing a reference for developing reasonable operating procedures and optimizing unit operation, ultimately achieving refined management and economical operation of the power plant.

[0004] However, some power plants currently have limited data processing and analysis capabilities, failing to fully tap into the information contained within the data. For example, they simply set alarm thresholds, triggering an alarm when the wall temperature exceeds the threshold. However, they lack in-depth analysis of temperature trends and correlations between different measurement points, making it difficult to predict potential faults in advance. Furthermore, data processing systems are not intelligent enough to automatically identify complex operating conditions and fault patterns, requiring extensive manual data screening and analysis. This is not only inefficient but also prone to misjudgments or omissions.

[0005] Therefore, a boiler wall temperature anomaly detection and early warning method based on sample entropy calculation is needed to improve the accuracy and robustness of anomaly detection and reduce false alarms and missed alarms. Summary of the Invention

[0006] The purpose of the present invention is to propose a method for detecting and warning abnormal boiler wall temperature based on sample entropy calculation, which includes the following steps:

[0007] Step 1: Divide the sliding window and use the sliding window to observe the temperature time series data;

[0008] Step 2: Calculate the statistics within each sliding window;

[0009] Step 3: Construct parameter matrix based on statistics;

[0010] Step 4: Based on the parameter matrix, use the OCSVM algorithm to perform anomaly detection and early warning.

[0011] Furthermore, the statistics in step 2 include kurtosis, skewness, and sample entropy.

[0012] Furthermore, the parameter matrix is:

[0013]

[0014] Among them, m is the number of sliding windows, K p is the kurtosis of the p-th sliding window, S p is the skewness of the p-th sliding window, H p is the sample entropy of the p-th sliding window, p = 1, 2, …, m.

[0015] Furthermore, the calculation formula of kurtosis is:

[0016]

[0017] Among them, K k is the kurtosis of the kth sliding window, w is the length of the sliding window, s is the sliding step size, i is the number of the length of the sliding window, T (k-1)s+i is the (k-1)s+ith data value in the sliding window, is the mean of the data in the sliding window.

[0018] Furthermore, the calculation formula for skewness is:

[0019]

[0020] Among them, S k is the skewness of the kth sliding window, w is the length of the sliding window, s is the sliding step size, i is the number of the length of the sliding window, T (k-1)s+i is the (k-1)s+ith data value in the sliding window, is the mean of the data in the sliding window.

[0021] Furthermore, step 4 specifically includes the following sub-steps:

[0022] Step 41: performing normalization preprocessing on the data samples in the parameter matrix;

[0023] Step 42: Use the normalized data to train the OCSVM model;

[0024] Step 43: Input each data sample in the parameter matrix into the trained OCSVM model for calculation to obtain the decision function value; if the decision function value is greater than the set threshold, the sample is determined to be a normal sample; if the decision function value is less than or equal to the set threshold, the sample is determined to be an abnormal sample, and an alert is issued that abnormal temperature fluctuations occur in the sliding window associated with the sample.

[0025] The beneficial effects of the present invention are:

[0026] 1. The sample entropy in the statistic of the present invention is sensitive to small changes in the data, can effectively capture the nonlinear characteristics and dynamic changes in the data, and provide characteristic information reflecting the inherent laws of the data for anomaly detection.

[0027] 2. The present invention utilizes the One-Class Support Vector Machine (OCSVM) algorithm, which does not require labeled data. It can use a small amount of normal samples to build a model, map the data into a high-dimensional space, and separate normal data from abnormal data by finding the optimal hyperplane. It has a good detection capability for unknown abnormal types.

[0028] 3. The features provided by the sample entropy in the present invention can provide OCSVM with more representative input, enabling OCSVM to better learn the distribution of normal data, improve the detection accuracy and robustness of abnormal boiler wall temperature data, and effectively detect various types of boiler wall temperature anomalies. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 The present invention is a flow chart of a method for detecting and warning abnormal boiler wall temperature based on sample entropy calculation.

[0030] Figure 2 This is a time series data diagram of the boiler wall temperature in a power plant.

[0031] Figure 3 It is a statistical characteristic indicator of boiler wall temperature.

[0032] Figure 4 This is the decision function value and threshold distribution diagram of the OCSVM anomaly detection algorithm.

[0033] Figure 5 The distribution map of outliers (red dots) detected using the OCSVM anomaly detection algorithm. DETAILED DESCRIPTION

[0034] The present invention proposes a method for detecting and warning abnormal boiler wall temperature based on sample entropy calculation. The present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0035] Figure 1 This is a flow chart of the method for detecting and warning abnormal boiler wall temperature based on sample entropy calculation in the present invention, which specifically includes:

[0036] Step 1: Sliding window division and data preparation;

[0037] Use sliding windows to observe temperature time series data. The data collection interval is Δt = 60 seconds. The number of sliding windows is m, set m = 1000; the length of the sliding window is w, set w = 300; the sliding step of the sliding window is Δt × s (s ≥ 1), set s = 60, such as Figure 2 shown.

[0038] Assume that the initial time is t0, and the temperature time series data is T={T1,T2,…,T n}(n≥(m-1)s+w). Among them, T z Indicates that at time t z =t0+(z-1)Δt(z=1,2,…,n) the collected temperature value.

[0039] The data W in the kth (k=1,2,…,m) sliding window k ={T (k-1)s+1 ,T (k-1)s+2 ,…,T (k-1)s+w}, the time interval is [t0+((k-1)s)Δt,t0+((k-1)s+w-1)Δt].

[0040] Step 2: The three statistics within each sliding window are: Kurtosis: measures the sharpness of the data distribution, usually expressed as K; Skewness: measures the asymmetry of the data distribution, usually expressed as S. Sample Entropy: measures the complexity of the time series, usually expressed as H. Calculate the kurtosis K of the data in the kth (k = 1, 2, ..., m) sliding window. k , skewness S k and sample entropy H k The kurtosis and skewness are calculated using the following formula:

[0041] Kurtosis of the kth sliding window:

[0042]

[0043] Among them, K k is the kurtosis of the kth sliding window, w is the length of the sliding window, s is the sliding step size, i is the number of the length of the sliding window, T (k-1)s+i is the (k-1)s+ith data value in the sliding window, is the mean of the data in the sliding window.

[0044] The skewness of the k-th sliding window:

[0045]

[0046] Among them, S k is the skewness of the kth sliding window, w is the length of the sliding window, s is the sliding step size, i is the number of the length of the sliding window, T (k-1)s+i is the (k-1)s+ith data value in the sliding window, is the mean of the data in the sliding window.

[0047] Step 3: Construct parameter matrix;

[0048] Combine the kurtosis, skewness, and sample entropy of the m windows into an m×3 matrix M.

[0049]

[0050] Set m = 1000, Figure 3 It is a statistical characteristic indicator of boiler wall temperature.

[0051] Step 4: Anomaly detection and warning based on the One-Class Support Vector Machine (OCSVM) algorithm;

[0052] 4.1 Data Preprocessing

[0053] Each row of the m×3 matrix M represents a data sample, and each column represents a different feature. There are m samples in total, each with three features. First, the data is normalized to ensure that different features have similar scales, preventing any one feature from having a significant impact on the results.

[0054] 4.2 Model Training

[0055] Use the prepared data to train the OCSVM model. The goal of OCSVM is to find an optimal hyperplane that contains as many normal samples as possible on one side of the hyperplane while excluding abnormal samples on the other side. During training, the model learns the normal distribution of the data, maps the data into a high-dimensional feature space, and calculates the similarity between data points using a kernel function.

[0056] 4.3 Anomaly Detection and Early Warning

[0057] One-Class Support Vector Machine (OCSVM) anomaly detection algorithm is used to perform anomaly detection on m points.

[0058] For the trained OCSVM model, each data sample in the matrix M is input into the model for calculation. The model calculates a decision function value for each sample, which represents the distance between the sample and the hyperplane.

[0059] After the model is trained, for each data sample in the matrix M, it is input into the OCSVM model, and the model outputs a decision function value. Figure 4 This is the decision function value and threshold distribution diagram of the OCSVM anomaly detection algorithm.

[0060] Set the threshold θ = -1, if the decision function value > θ, then the sample is judged as a normal sample; if the decision function value ≤ θ, then the sample is judged as an abnormal sample. k ,S k ,H k ) is ≤θ, the point is determined to be an abnormal point, and an alert is issued for abnormal temperature fluctuations within the start and end time periods of the sliding window associated with the point [t0+((k-1)s)Δt, t0+((k-1)s+w-1)Δt]. Where k∈[1,2,…,m]. Figure 5 The distribution map of outliers (red dots) detected using the OCSVM anomaly detection algorithm.

[0061] In summary, the present invention can improve the detection accuracy and robustness of boiler wall temperature anomaly data and effectively detect various types of boiler wall temperature anomalies.

Claims

1. A method for detecting and warning abnormal boiler wall temperature based on sample entropy calculation, characterized in that: The following steps are involved: Step 1: Divide the sliding window and use the sliding window to observe the temperature time series data; Step 2: Calculate the statistics within each sliding window; Step 3: Construct parameter matrix based on statistics; Step 4: Based on the parameter matrix, use the OCSVM algorithm to perform anomaly detection and early warning.

2. The method for detecting and warning abnormal boiler wall temperature based on sample entropy calculation according to claim 1 is characterized in that: The statistics in step 2 include kurtosis, skewness, and sample entropy.

3. The method for detecting and warning abnormal boiler wall temperature based on sample entropy calculation according to claim 2 is characterized in that: The parameter matrix is: Among them, m is the number of sliding windows, K p is the kurtosis of the p-th sliding window, S p is the skewness of the p-th sliding window, H p is the sample entropy of the p-th sliding window, p = 1, 2, …, m.

4. The method for detecting and warning abnormal boiler wall temperature based on sample entropy calculation according to claim 2 is characterized in that: The calculation formula of the kurtosis is: Among them, K k is the kurtosis of the kth sliding window, w is the length of the sliding window, s is the sliding step size, i is the number of the length of the sliding window, T (k-1)s+i is the (k-1)s+ith data value in the sliding window, is the mean of the data in the sliding window.

5. The method for detecting and warning abnormal boiler wall temperature based on sample entropy calculation according to claim 2 is characterized in that: The calculation formula of the skewness is: Among them, S k is the skewness of the kth sliding window, w is the length of the sliding window, s is the sliding step size, i is the number of the length of the sliding window, T (k-1)s+i is the (k-1)s+ith data value in the sliding window, is the mean of the data in the sliding window.

6. The method for detecting and warning abnormal boiler wall temperature based on sample entropy calculation according to claim 1 is characterized in that: Said step 4 specifically includes the following sub-steps: Step 41: performing normalization preprocessing on the data samples in the parameter matrix; Step 42: Use the normalized data to train the OCSVM model; Step 43: Input each data sample in the parameter matrix into the trained OCSVM model for calculation to obtain the decision function value; If the decision function value is greater than the set threshold, the sample is judged to be a normal sample; If the decision function value is less than or equal to the set threshold, the sample is determined to be an abnormal sample, and an alert is issued that abnormal temperature fluctuations occur in the sliding window associated with the sample.

Images