A method for predicting and warning abnormal operating behavior of hydro turbine units

Abstract

This invention discloses a method for predicting and warning of abnormal operating behavior of hydro-turbine units. The degree of abnormality can be used to determine the current operating status of the hydro-turbine unit, and the abnormal data can be intuitively seen as abnormal monitoring quantities. This method can realize the monitoring and early warning of abnormal operating behavior of hydro-turbine units, support the intelligent management of hydropower plants, improve the operating efficiency of power plants, reduce operating costs, and enhance the well-being of power plant staff.

Description

Technical Field

[0001] This invention relates to the field of power technology, and in particular to a method for predicting and warning of abnormal operating behavior of hydro turbine units. Background Technology

[0002] Traditional methods for determining abnormal conditions in hydroelectric turbine units rely on the normality of measurements taken at key monitoring points. For individual monitoring points, a "threshold setting" is typically used to assess abnormalities. If the data falls within the specified threshold, it's considered normal; otherwise, it's deemed abnormal. This method is only suitable for confirming and locating faults after they occur, making it difficult to provide early warnings in the early stages of a fault. Furthermore, threshold over-limit alarms only apply to a single monitoring parameter, neglecting the correlation between parameters. To prevent false alarms, threshold ranges are generally set large, triggering alarms only when significant anomalies occur. This lack of flexibility in alarm handling can easily lead to irreparable losses. Summary of the Invention

[0003] The purpose of this invention is to provide a method for predicting and warning of abnormal operating behavior of a hydro-turbine unit in order to solve the above-mentioned problems. This invention utilizes a large amount of normal data accumulated during the unit's past normal operation to establish a predictive and early warning model for abnormal conditions of the hydro-turbine unit. It can realize real-time monitoring of the overall abnormal condition of the hydro-turbine unit and provide early warning when the unit's condition shows an abnormal trend, so as to prompt on-site personnel to take relevant measures in a timely manner to effectively protect the safe operation of the unit.

[0004] The present invention achieves the above objectives through the following technical solutions:

[0005] A method for predicting and warning of abnormal operating behavior of a hydro-turbine unit includes the following steps:

[0006] Step 1: Data preparation, including:

[0007] Data extraction: Select several typical monitoring quantities that can reflect the status of the turbine unit from the turbine unit monitoring system, and extract them into a data table according to the time period based on the monitoring quantities;

[0008] Historical data preprocessing: The extracted historical data of multiple monitoring quantities are processed for missing and duplicate values. Missing values ​​are filled in according to the previous values ​​of the data. If all values ​​of a data point are exactly the same as a data point in the data table, the data point is deleted.

[0009] Step 2: Building the data prediction model, including:

[0010] The model is constructed using a nearest neighbor-based prediction algorithm.

[0011] Assuming there are n monitoring parameters reflecting the turbine unit's status selected from step 1, and m data entries in the data table, then the dimension of the real-time data reflecting the turbine unit's status is n, which can be represented as [x1 x2 … x n Historical data for the turbine units are represented as follows: The predictive model algorithm works by calculating the distance between each data entry in the historical data table and the real-time data, such as the distance L between the real-time data and the j-th data entry in the historical data table. j As in equation (1):

[0012]

[0013] The smallest of the m calculated distances is used as the predicted value of the real-time operating status data of the turbine unit, and the value of this distance is defined as the anomaly degree of the operating status of the turbine unit.

[0014] Step 3: Early warning of abnormal behavior of the turbine unit, specifically including:

[0015] Anomaly sequence construction: Combining the unit operation pattern, set the recent appropriate time length (such as the past 6 months) of turbine unit status data and the previous historical turbine unit data, and substitute them into the data prediction model constructed in step 2 to obtain the anomaly degree of each turbine unit status data within the set time range.

[0016] The Pauta criterion is used to construct an early warning threshold for the anomaly degree of real-time turbine units. The Pauta criterion is used to calculate the anomaly degree sequence to obtain the judgment standard value of the anomaly degree corresponding to the real-time collected data. If the value of the anomaly degree of real-time turbine units is greater than the judgment standard value, the current turbine unit status is judged as abnormal, otherwise it is judged as healthy.

[0017] Early warning analysis and alarm output; if the real-time turbine unit anomaly level is determined to be healthy, the calculation process ends and the data is stored; if the real-time turbine unit anomaly level is determined to be abnormal, the monitoring quantities in the current turbine unit status data that deviate significantly from the predicted values ​​are calculated to analyze and obtain the monitoring quantities that need attention, as follows:

[0018] If the predicted value of the real-time operating data of the turbine unit is the j-th data in the historical data table, then the deviation value of the ith monitoring quantity is as shown in equation (2):

[0019] Δ ji =(y ji -x i ) 2 , where 1≤j≤m, 1≤i≤n (2)

[0020] Calculate the deviation values ​​of n monitoring quantities, and select the monitoring quantities whose deviation values ​​are greater than the average deviation value as the monitoring quantities that need to be monitored.

[0021] A further proposed approach is that, in step 1, typical monitoring quantities include turbine head, active power and reactive power, bearing temperature, stator and rotor temperature, technical water supply pressure and flow rate, excitation current, cooling water temperature, and other monitoring quantities.

[0022] A further proposed solution is that, in step 3, the warning information is issued by including the time, the monitoring quantity that needs attention, and the corresponding deviation value.

[0023] The beneficial effects of this invention are as follows:

[0024] This invention provides a method for predicting and warning of abnormal operating behavior of a hydro-turbine unit. The degree of abnormality can be used to determine the current operating status of the hydro-turbine unit, and the abnormal data can be intuitively identified. This method can monitor and warn of abnormalities in the operating behavior of the hydro-turbine unit, support the intelligent management of hydropower plants, improve the operating efficiency of power plants, reduce operating costs, and enhance the well-being of power plant staff.

[0025] This invention utilizes the massive amount of status information accumulated during the normal operation of the turbine unit and uses historical database operation data to predict the current operating status of the turbine unit. The computational complexity is not high, and the method is practical and feasible. Attached Figure Description

[0026] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0027] Figure 1 This is a flowchart of the present invention. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be described in detail below. Obviously, the described embodiments are merely some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other implementation methods obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0029] In any embodiment, such as Figure 1 As shown, the present invention provides a method for predicting and warning of abnormal operating behavior of a hydro-turbine unit. Taking the operating data of the 02F hydro-turbine generator of a power plant as an example, the invention will be further described in detail below:

[0030] This invention enables real-time identification and analysis of abnormal operating states of hydro-turbine units. First, data extraction and preprocessing are performed to prepare computational data. Then, a predictive model is built based on the data to calculate the degree of abnormality in unit operation. Next, the Pauta criterion is used to construct early warning thresholds to achieve early warning of hydro-turbine unit anomalies. Finally, the abnormal data is analyzed to obtain detailed early warning information, thereby outputting alarms. The identification process is as follows: Figure 1 As shown, the details are as follows:

[0031] First, data preparation, including data extraction and preprocessing, was performed. Based on the operational data collection of a mixed-flow turbine unit, 17 monitoring parameters were extracted at 10-minute intervals, including turbine head, active and reactive power, bearing temperature, stator and rotor temperature, technical water supply pressure and flow rate, excitation current, and cooling water temperature. Missing values ​​for these parameters were supplemented with previous values, and duplicate data were deleted. Taking the data from 23:00 to 23:30 on December 30, 2022, as an example, the processed data is shown in Table 1.

[0032] Table 1. Data after preprocessing

[0033]

[0034]

[0035] Secondly, a predictive model was constructed. Seventeen monitoring parameters reflecting the turbine unit's status were selected. Using 50,368 data points from January 1, 2022 to December 31, 2022 as historical data, a model was built to predict real-time data from January 1, 2023 to February 27, 2023. Taking the data at 23:50 on February 27, 2023 as an example, the dimensions of the real-time data reflecting the turbine unit's status are 17, which can be represented as [x1 x2 … x 17 Historical data for the turbine units are as follows: The distance between each data point in the historical data and the real-time data is calculated, as shown in Table 2.

[0036] Table 2 Anomalies of Real-Time and Historical Data

[0037] Data sequence number 1 2 3 4 …… 50366 50367 50368 Anomaly 2125.15 2125 2125 2125 …… 50 84 2125

[0038] The data with the lowest anomaly is the data at 8:30 on September 3, 2022. The predicted values ​​of the turbine unit's operating status are shown in Table 3. The anomaly of the current turbine unit's operating status is 15.

[0039] Table 3 Comparison of Real-time Data and Predicted Data

[0040]

[0041]

[0042] Finally, early warnings were issued for abnormal unit behaviors from January 1, 2023 to February 27, 2023, including the construction of anomaly sequences, the construction of early warning thresholds, early warning analysis, and alarm output.

[0043] 1) Construction of anomaly sequence. The turbine unit status data from January 1, 2023 to February 27, 2023 and the historical data of the turbine unit before January 1, 2023 were selected and substituted into equation (1) to calculate the anomaly sequence.

[0044] 2) Early warning threshold construction. The Pauta criterion was used to calculate the anomaly series from January 1, 2023 to February 27, 2023, and the threshold was 36. The early warning results based on the early warning threshold are shown in Table 4.

[0045] Table 4. Early Warning Analysis of Abnormal Behavior of Hydropower Units

[0046] time Anomaly Warning threshold Warning results 2023-1-20 23:40 44 36 abnormal 2023-1-20 18:40 29 36 normal

[0047] 3) Early warning analysis and alarm output. In Table 4, the data at 18:40 on January 20, 2023 is judged as normal and the corresponding data is stored; the data at 23:40 on January 20, 2023 is judged as abnormal. The turbine unit operation data at that time is substituted into formula (2) to obtain the deviation values ​​of each monitoring quantity. The average deviation value is 7.22. The monitoring quantities that need attention are calculated and analyzed as shown in Table 5.

[0048] Table 5 Analysis of Monitoring Quantities Requiring Attention

[0049] Monitoring volume Deviation value Unit technical water supply main flow rate 21.49 Main transformer technology water supply main pipe flow 19.65 Water-conducting tile temperature 15.67 Stator winding temperature 17.10 Stator core temperature 18.36 Temperature at the upper guide bearing 10.50 Temperature at the lower guide bearing 9.30

[0050] Therefore, the warning message is as follows:

[0051] "An anomaly was detected in the data at 23:40 on January 20, 2023. The monitoring quantities and deviations that need attention are: 'Unit technical water supply main pipe flow rate': 21.49; 'Main transformer technical water supply main pipe flow rate': 19.65; 'Water guide bearing temperature': 15.67; 'Stator winding temperature': 17.10; 'Stator core temperature': 18.36; 'Upper guide bearing temperature': 10.50; 'Lower guide bearing temperature': 9.30."

[0052] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims. It should also be noted that the various specific technical features described in the above embodiments can be combined in any suitable manner without contradiction. To avoid unnecessary repetition, the present invention will not describe the various possible combinations separately. Furthermore, various different embodiments of the present invention can also be arbitrarily combined, as long as they do not violate the spirit of the present invention, they should also be considered as the content disclosed in the present invention.

Claims

1. A method for predicting and warning of abnormal operating behavior of a hydro-turbine unit, characterized in that, Includes the following steps: Step 1: Data preparation, including: Data extraction: Select several typical monitoring quantities that can reflect the status of the turbine unit from the turbine unit monitoring system, and extract them into a data table according to the time period based on the monitoring quantities; Historical data preprocessing: The extracted historical data of multiple monitoring quantities are processed for missing and duplicate values. Missing values ​​are filled in according to the previous values ​​of the data. If all values ​​of a data point are exactly the same as a data point in the data table, the data point is deleted. Step 2: Building the data prediction model, including: The model is constructed using a nearest neighbor-based prediction algorithm. Assuming there are n monitoring parameters reflecting the turbine unit's status selected from step 1, and m data entries in the data table, then the dimension of the real-time data reflecting the turbine unit's status is n, which can be represented as [x1 x2 … x n Historical data for the turbine units are represented as follows: The predictive model algorithm works by calculating the distance between each data entry in the historical data table and the real-time data, such as the distance L between the real-time data and the j-th data entry in the historical data table. j As in equation (1): The smallest of the m calculated distances is used as the predicted value of the real-time operating status data of the turbine unit, and the value of this distance is defined as the anomaly degree of the operating status of the turbine unit. Step 3: Early warning of abnormal behavior of the turbine unit, specifically including: Anomaly sequence construction: Combining the unit operation pattern, the recent turbine state data and previous historical turbine data are substituted into the data prediction model constructed in step 2 to obtain the anomaly of each turbine state data within the set time range. The Pauta criterion is used to construct an early warning threshold for the anomaly degree of real-time turbine units. The Pauta criterion is used to calculate the anomaly degree sequence to obtain the judgment standard value of the anomaly degree corresponding to the real-time collected data. If the value of the anomaly degree of real-time turbine units is greater than the judgment standard value, the current turbine unit status is judged as abnormal, otherwise it is judged as healthy. Early warning analysis and alarm output; if the real-time turbine unit anomaly level is determined to be healthy, the calculation process ends and the data is stored; if the real-time turbine unit anomaly level is determined to be abnormal, the monitoring quantities in the current turbine unit status data that deviate significantly from the predicted values ​​are calculated to analyze and obtain the monitoring quantities that need attention, as follows: If the predicted value of the real-time operating data of the turbine unit is the j-th data in the historical data table, then the deviation value of the i-th monitoring quantity is as shown in equation (2): D ji =(y ji -x i ) 2 ,wherein1≤j≤m,1≤i≤n (2) Calculate the deviation values ​​of n monitoring quantities, and select the monitoring quantities whose deviation values ​​are greater than the average deviation value as the monitoring quantities that need to be monitored.

2. The method for predicting and warning of abnormal operating behavior of a hydro-turbine unit as described in claim 1, characterized in that, In step 1, typical monitoring quantities include turbine head, active power and reactive power, bearing temperature, stator and rotor temperature, technical water supply pressure and flow rate, excitation current, cooling water temperature, etc.

3. The method for predicting and warning of abnormal operating behavior of a hydro-turbine unit as described in claim 1, characterized in that, In step 3, the early warning information is issued by including the time, the monitoring quantity that needs attention, and the corresponding deviation value.

Images