A power plant equipment early warning method based on similarity

Abstract

The application discloses a power plant equipment early warning method based on similarity, which calculates the similarity between the current important parameters of power plant equipment and the historical normal state database, discovers the equipment abnormality in advance, and reduces the failure rate of power plant equipment. The application considers the coupling relationship between parameters by using the similarity calculation method, measures the health degree of equipment by calculating the similarity of distance information and direction information, and thus performs the failure early warning of power plant equipment.

Description

Technical Field

[0001] This invention relates to a method for early warning of power plant equipment faults, specifically a method for early warning of power plant equipment faults based on similarity. Background Technology

[0002] Power plant equipment typically operates for over 10 years. As operating time increases and some equipment operates in harsh environments, equipment failures can easily lead to significant safety accidents and economic losses. Therefore, early warning functions are necessary. However, current popular early warning methods are somewhat outdated, usually involving setting fixed thresholds for key equipment parameters. When a parameter exceeds the threshold, an early warning is issued. However, key equipment parameters are generally coupled, and even if all parameters are within the threshold range, it does not necessarily mean that the equipment is not malfunctioning in its current operating state. This invention uses a similarity calculation method to take into account the coupling relationships between parameters.

[0003] Typically, a fixed threshold is set for the important characteristic parameters of the equipment, and an early warning is issued when a parameter exceeds the threshold. However, there is usually a coupling relationship between the important parameters of the equipment. Even if all parameters are within the threshold range, it does not mean that the equipment has not malfunctioned in the current operating state. Summary of the Invention

[0004] The purpose of this invention is to provide a power plant equipment early warning method based on similarity. This method uses a similarity calculation method to take into account the coupling relationship between parameters. It measures the health status of the equipment by calculating the similarity of distance information and direction information, thereby providing early warning of faults for power plant equipment.

[0005] The present invention is achieved using the following technical solution:

[0006] A similarity-based early warning method for power plant equipment is proposed. This method calculates the similarity between the current important parameters of power plant equipment and a historical normal status database to detect equipment anomalies in advance and reduce the failure rate of power plant equipment.

[0007] A further improvement of the present invention is that it specifically includes the following steps:

[0008] Step 1: Collect historical data, collect key characteristic data of early warning equipment, select stable operating condition data, and obtain a stable operating condition dataset;

[0009] Step 2: Based on the stable operating condition dataset from Step 1, clean the abnormal data;

[0010] Step 3: Based on the cleaning data obtained in Step 2, select data within the similar normal range according to the key characteristic data of the unit;

[0011] Step 4: Determine whether the current operating state is a stable operating condition. If it is a stable operating condition, calculate the similarity values between the current characteristic values of the warning device and each piece of data in Step 3, and select the maximum value among the similarity values. If this value is less than the set threshold, start the warning.

[0012] A further improvement of the present invention is that in Step 1, to collect the important characteristic data of the warning device, the required measuring points are shown in Table 1:

[0013] Table 1. Measuring point data

[0014] Serial Number name symbol unit Remark 1 Feature 1 x1 range of change 2 Feature 2 x2 3 Feature n xn

[0015] A further improvement of the present invention is that in Step 2, according to the historical data collected in Table 1, use the box plot principle to remove the missing values and outliers of each characteristic, that is, the values not within the range [Q1 - 1.5IQR, Q3 + 1.5IQR] and all characteristic values at their corresponding timestamps.

[0016] A further improvement of the present invention is that in Step 3, select the similar data within the normal range from Step 2. To ensure that the comparison database is not too complex when calculating the similarity in Step 4 and to ensure that the calculation can be completed in a short time to achieve the purpose of rapid warning, it is necessary to screen out all the characteristic values at the corresponding timestamps according to the floating range of the current value of the specified characteristic value to obtain the comparison database. When the change of the specified characteristic value exceeds 5%, re-screen the comparison database and perform standardization processing on the data.

[0017] A further improvement of the present invention is that in Step 3, the data is standardized according to the following formula:

[0018]

[0019] where μ is the mean of all sample data and σ is the standard deviation of all sample data.

[0020] A further improvement of the present invention is that in Step 4, calculate the similarity values between the current characteristic values of the device and each piece of data in the comparison database, select the maximum value HMAX among the similarity values. If HMAX < H, where H is a pre-given threshold, start the warning.

[0021] A further improvement of the present invention is that in Step 4, the specific similarity calculation method is as follows:

[0022] Use the set model method based on distance and direction information to describe the similarity of the device. The current stable operating condition data c q and the historical operating data c i The similarity description is:

[0023]

[0024]

[0025]

[0026] In the formula D(c q ,c i ) represents distance information, x qj and x ij Let cos(δ) represent the j-th feature value of the current operating condition data and the j-th feature value of the i-th data in the comparison database, respectively; i ) represents directional information; w1 and w2 are weighting factors, and n is the number of important device features in the selected area; where

[0027]

[0028]

[0029]

[0030] The similarity between the equipment's current operating conditions and historical normal operating conditions is calculated using equation (1.1). When the maximum similarity value is less than a manually given threshold S... d When the equipment is considered to be malfunctioning, an early warning alarm is issued, and the greater the similarity, the healthier the equipment is.

[0031] A further improvement of the present invention is that w1 and w2 are respectively 0.8 and 0.2.

[0032] The present invention has at least the following beneficial technical effects:

[0033] First, by identifying stable operating conditions, data from when the equipment is running stably is obtained, thus excluding data from when the equipment is accelerating or decelerating. Then, outlier data is eliminated using the box plot outlier detection method. To optimize program execution time, data similar to a certain key feature (such as power) is found, and the data is standardized to form a comparison database. Finally, the similarity values ​​between the data are calculated using directional and distance information, as shown in Formula 1.2 and Table 4. This can, to a certain extent, represent the health status of the equipment, thereby providing early warnings for the equipment. Attached Figure Description

[0034] Figure 1 This is a flowchart of a power plant equipment early warning method based on similarity according to the present invention. Detailed Implementation

[0035] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided so that the present disclosure can be more thoroughly understood and the scope of the present disclosure can be fully conveyed to those skilled in the art. It should be noted that, without conflict, the embodiments in the present invention and the features in the embodiments can be combined with each other. The present invention will be described in detail below with reference to the drawings and in conjunction with the embodiments.

[0036] As Figure 1 shown, a method for early warning of power plant equipment based on similarity provided by the present invention includes the following steps:

[0037] 1. Collect historical data, that is, key feature data of the early warning equipment. The required measuring points are shown in Table 1. Select stable operating condition data, give the change range of one or several key features within 5 minutes, and retain all feature data with the same time stamp accordingly.

[0038] Table 1. Measuring point data

[0039] Serial Number name symbol unit Remark 1 Feature 1 x1 range of change 2 Feature 2 x2 3 Feature n xn

[0040] 2. Data cleaning. Remove the missing values and outliers of each feature according to the historical data collected in Table 1 using the principle of box plot, that is, the values not within the range [Q1 - 1.5IQR, Q3 + 1.5IQR] and all feature values under their corresponding time stamps.

[0041] 3. Select similar data. Select similar data within the normal range from step 2). To make the comparison database not too complex when calculating the similarity in step 4) and ensure that the calculation can be completed in a short time to achieve the purpose of rapid early warning, it is necessary to screen all feature values with the corresponding time stamp according to the floating range (such as 20%) of the current value of the specified feature value (such as power) to obtain the comparison database. When the change of the specified feature value exceeds 5%, it is necessary to re-screen the comparison database and standardize the data according to the following formula:

[0042]

[0043] where μ is the mean of all sample data and σ is the standard deviation of all sample data.

[0044] 4. Calculate the similarity value between the current feature value of the equipment and each piece of data in the comparison database, and select the maximum value HMAX of the similarity values. If HMAX < H (a pre-given threshold), then start early warning.

[0045] The specific similarity calculation method is as follows:

[0046] The similarity description of the devices is performed using an ensemble model method based on distance and orientation information, with the current stable operating condition data c. q and historical running data c i The similarity can be described as:

[0047]

[0048]

[0049]

[0050] In the formula D(c q ,c i ) represents distance information, x qj and x ij Let represent the j-th feature value of the current operating condition data and the j-th feature value of the i-th data in the comparison database, respectively. cos(δ) i The ) represents directional information. w1 and w2 are weighting factors, which can be selected according to the actual situation, generally 0.8 and 0.2 respectively. n is the number of important features of the device in the selected area.

[0051]

[0052]

[0053]

[0054] The similarity between the equipment's current operating conditions and historical normal operating conditions is calculated using equation (1.1). When the maximum similarity value is less than a manually given threshold S... d When the equipment is considered to be malfunctioning, an early warning alarm is issued, and the greater the similarity, the healthier the equipment is.

[0055] Example

[0056] Taking a coal mill in a power plant as an example, the details are as follows:

[0057] 1. Collect historical data, i.e., key characteristic data of the early warning equipment. The required measurement points are shown in Table 2. Select stable operating condition data, given that the change in motor power characteristics within 15 minutes is less than 150, and retain all characteristic data with the same timestamp, as shown in Table 3.

[0058] Table 2 Key Feature Data of Early Warning Equipment

[0059]

[0060] Table 3 Stable operating condition data

[0061]

[0062]

[0063] 2. Data cleaning: Based on the historical data of stable operating conditions in Table 3, use the principle of box plot to remove the missing values and outliers of each feature, that is, the values not within the range [Q1 - 1.5IQR, Q3 + 1.5IQR] and all feature values at their corresponding timestamps. The remaining clean data is shown in Table 4.

[0064] Table 4 Clean data

[0065]

[0066] 3. Select similar data: Select similar data within the normal range from step 2). To make the comparison database not too complex when calculating the similarity in step 4) and ensure that the calculation can be completed in a short time to achieve the purpose of rapid warning, it is necessary to screen out all feature values at the corresponding timestamps according to the floating range (such as when the current power value is 502, select the power range of 452 - 552 according to experience) of the current value of the specified feature (such as power), obtain the comparison database, and standardize the data according to the following formula:

[0067] The standardized data is shown in Table 5.

[0068] Table 5 Standardized comparison database

[0069]

[0070] 4. Calculate the similarity value between the current feature value of the device and each piece of data in the comparison database, and select the maximum value HMAX of the similarity values. If HMAX < H (a pre-given threshold, which can be taken as 0.8 according to experience), then start warning. Take a current normal data and assume an abnormal data as shown in Table 6. After standardizing the data according to the mean and variance in the comparison database, it is shown in Table 7

[0071] Table 6 Current data of the device

[0072]

[0073] Table 7 Standardized current data of the device

[0074]

[0075] Calculate the similarity between Table 7 and the first piece of data in Table 5:

[0076]

[0077]

[0078]

[0079]

[0080]

[0081]

[0082] The similarity between the current normal data in Table 6 and the first historical data in Table 5 is 0.126.

[0083] Using this method, the similarity between normal data, abnormal data, and each data point in the comparison database was calculated. The maximum similarity value, 75th percentile, 50th percentile, and 25th percentile were obtained, as shown in Table 8: 0.945, 0.642, 0.472, and 0.311, respectively. Since the maximum similarity value of abnormal data, HMAX = 0.563 < 0.8, an alarm needs to be triggered.

[0084] Table 8 Similarity Values

[0085]

[0086] Although the present invention has been described in detail above with general descriptions and specific embodiments, modifications or improvements can be made to it, which will be obvious to those skilled in the art. Therefore, all such modifications or improvements made without departing from the spirit of the present invention fall within the scope of protection claimed by the present invention.

Claims

1. A power plant equipment early warning method based on similarity, characterized in that, This method calculates the similarity between the current important parameters of power plant equipment and the historical normal state database to detect equipment anomalies in advance and reduce the failure rate of power plant equipment. It specifically includes the following steps: Step 1: Collect historical data, collect important characteristic data of early warning equipment, and select stable operating condition data to obtain a stable operating condition data set. Step 2: Based on the stable operating condition data set in Step 1, clean the abnormal data. Step 3: Based on the cleaned data obtained in Step 2, select data within a similar normal range according to the key characteristic data of the unit; select data within a similar normal range from Step 2. To ensure that the comparison database is not too complex when calculating the similarity in Step 4 and to ensure that the calculation is completed in a short time to achieve the purpose of rapid early warning, all characteristic values corresponding to the time stamp need to be screened according to the floating range above and below the current value of the specified characteristic value to obtain a comparison database. When the change of the specified characteristic value exceeds 5%, the comparison database is re-screened and the data is standardized. Step 4: Determine whether the current operating state is a stable operating condition. If it is a stable operating condition, calculate the similarity value between the current characteristic value of the early warning equipment and each piece of data in Step 3, and select the maximum value of the similarity values. If this value is less than the set threshold, start early warning; calculate the similarity value between the current characteristic value of the equipment and each piece of data in the comparison database, and select the maximum value HMAX of the similarity values. If HMAX < H, where H is a pre-given threshold, start early warning. The specific similarity calculation method is as follows: Device similarity description is performed using an ensemble model method based on distance and orientation information, based on current stable operating condition data. c q and historical operating data c i The similarity is described as follows: (1.2) (1.3) (1.4) In the formula D ( c q , c i () represents distance information. x qj and x ij Let cos(δ) represent the j-th feature value of the current operating condition data and the j-th feature value of the i-th data in the comparison database, respectively; i ) indicates direction information; w 1 and w 2 is the weighting factor, and n is the number of important device features in the selected area; where (1.5) (1.6) (1.7) The similarity between the equipment's current operating conditions and historical normal operating conditions is calculated using equation (1.1). When the maximum similarity value is less than a manually given threshold S... d When the equipment is considered to be malfunctioning, an early warning alarm is issued, and the greater the similarity, the healthier the equipment is.

2. The power plant equipment early warning method based on similarity according to claim 1, characterized in that, In Step 2, according to the collected historical data, use the box plot principle to remove the missing values and abnormal values of each characteristic, that is, the values not within the range [Q1 - 1.5IQR, Q3 + 1.5IQR] and all characteristic values under their corresponding time stamps.

3. A power plant equipment early warning method based on similarity according to claim 1, characterized in that, In Step 3, the data is standardized according to the following formula: (1.1) where μ is the mean of all sample data, and σ is the standard deviation of all sample data.

4. The power plant equipment early warning method based on similarity according to claim 1, characterized in that, w 1 and w 2. Take values ​​of 0.8 and 0.2 respectively.

Images