An abnormality identification and early warning system and method for a hydroelectric generating unit

Abstract

The application discloses an abnormality identification early warning system of a hydroelectric generating set, comprising a history database, a peak value database, a searching module, a parameter database, a detecting module, an abnormality database, an early warning module and a processing module, and an abnormality identification early warning method of the hydroelectric generating set, which comprises the following steps: S1, periodically analyzing and calculating the operation state data of the generating set by using wavelet transform, and storing the peak value data into the peak value database; S2, performing abnormality detection on the peak value data in the peak value database, and storing the abnormality data into the abnormality database; and S3, transmitting the abnormality data to the processing module, and outputting a processing scheme.The application has the beneficial effect that the signal periodic analysis technology can be used to perform early warning on the abnormal operation state of the generating set without affecting the operation state of the generating set.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of unit state recognition and early warning, and particularly relates to an abnormality recognition and early warning system and method for a hydroelectric generating unit. BACKGROUND

[0002] A hydroelectric generating unit is a power device that converts the potential energy lost by water into mechanical energy, and the mechanical energy drives the rotor of the unit to rotate, and finally outputs alternating current. When the hydroelectric generating unit is running, the water wheel of the water turbine rotates under the action of the potential energy of water, and it converts the potential energy of water into the mechanical energy of the water turbine. Then, the unit rotates under the driving of the water wheel shaft to generate electricity, and finally outputs alternating current. Many important mechanical devices of the unit present a periodic operation law, such as the measured values of the gate opening, the oil level and the oil pressure.

[0003] In the prior art, the conventional detection and analysis method needs to manually statistically and analyze the related data such as the active power, the stator current, the stator core temperature, the cooling water inlet and outlet temperature, the cooling water flow and the cooling water pressure of the hydroelectric generating unit in the computer monitoring system intermittently, which is easily affected by the operation conditions of the generator and the technical experience of the statistical analysis personnel, and there is a problem that the abnormal state of the unit operation cannot be warned.

[0004] For example, a method for determining abnormality of single-machine active power closed-loop regulation function of a hydroelectric generator is disclosed in Chinese Patent Literature, with the publication number CN110989548A and the application date of November 1, 2019. The invention adopts four kinds of abnormality monitoring modules to comprehensively determine the abnormality of the single-machine active power closed-loop regulation function, which can effectively prevent the occurrence of missed judgment, but there is a problem that the abnormal state of the unit operation cannot be warned. SUMMARY

[0005] In view of the deficiency that the prior art cannot warn the abnormal state of the unit operation, the present application provides an abnormality recognition and early warning system and method for a hydroelectric generating unit, which can warn the abnormal state of the unit operation without affecting the operation state of the unit by using signal cycle analysis technology.

[0006] The technical solution of the present application is as follows: an abnormality recognition and early warning system for a hydroelectric generating unit, comprising: a historical database for saving historical operation state data of the unit;

[0007] a peak value database for saving peak value data obtained by periodically analyzing the historical operation state data of the unit;

[0008] a searching module for reading condition parameters for searching peak values from the parameter database, and then periodically analyzing the data in the historical database to obtain peak value data and store the peak value data in the peak value database, and connecting the historical database, the peak value database and the parameter database;

[0009] a parameter database for storing the limited condition parameters for peak searching;

[0010] a detection module for reading the peak data from the peak database, monitoring the peak data in the peak database in real time through sliding window and timing task, and storing the obtained abnormal data into an abnormal database, connecting the peak database and the abnormal database;

[0011] an abnormal database for storing the abnormal data of the detection module;

[0012] a warning module for reading the abnormal data and warning through a monitoring page, connecting the abnormal database and a processing module;

[0013] a processing module for judging the abnormal data and outputting solution suggestion information.

[0014] Preferably, the historical database stores the unit historical running state data in a time sequence format, and the unit historical running state data includes oil pressure data of an oil tank, oil level data of the oil tank, unit vibration amplitude data, and unit vibration direction data.

[0015] Preferably, the limited condition parameters are verified and uploaded to the parameter database, the limited condition parameters match the peak data of the unit historical running state data, and one limited condition parameter corresponds to one or more peak data.

[0016] Preferably, the peak database stores the rising rate, the falling rate, the rising time, and the falling time of the periodic signal.

[0017] Preferably, an abnormality identification and warning method for a hydroelectric unit includes the following steps:

[0018] S1: performing periodic analysis and calculation on the unit running state data by using wavelet transform, and storing the peak data into a peak database;

[0019] S2: performing abnormality detection on the peak data in the peak database, and storing the abnormal data into an abnormal database;

[0020] S3: transmitting the abnormal data to a processing module, and outputting a processing scheme.

[0021] Preferably, the wavelet transform is used to perform periodic analysis and calculation on the unit running state data, and the peak data is stored into the peak database, including the following steps:

[0022] S11: establishing a noise data processing model, performing linear interpolation and Gaussian smoothing filter processing on the running parameters of the unit obtained from the historical database;

[0023] S12: a peak value recognition model is established to find the peak value of the parameter information of the unit, and peak value data is obtained;

[0024] S13: the peak value data is analyzed and processed to obtain the rising rate, falling rate, rising time and falling time of the periodic signal and store them in the peak value database.

[0025] As preferred, the Gaussian smoothing filter expression is as follows:

[0026]

[0027] In the above formula, G(x, y) represents the Gaussian convolution kernel, (i, j) is the central value of the data, ΔE is the difference value at the positions of (i, j) and (x, y), σ is the standard deviation, and th is the threshold value.

[0028] After inputting the data, it is judged whether the parameter information in the neighborhood and the parameter information of the central point are greater than the threshold value, if yes, the weight value of this point is set to 0, otherwise, it is still the initial value;

[0029] The parameter point with a weight value of 0 is subjected to interpolation processing, and the values of the adjacent data points are used to do mean interpolation of the parameter point.

[0030] As preferred, a peak value recognition model is established to find the peak value of the parameter information of the unit, including the following steps:

[0031] S121: find the peak value in the data vector by wavelet transform, and use convolution to smooth the data vector;

[0032] S122: perform transposition difference processing on the processed data to obtain the difference data and the subscript array with a value of 0, perform difference processing on the array again to determine the level of the waveform, group the obtained array, and record the time information, and process the array of the horizontal waveform to make there be only one peak value or trough value in the range of each horizontal waveform;

[0033] S123: judge whether the prominence of the peak value or trough value is greater than the prominence threshold value and whether the width is greater than the width threshold value, if yes, the peak value or trough value meets the expectation.

[0034] As preferred, the peak value data in the peak value database is subjected to abnormality detection, and the abnormal data is stored in the abnormal database, including the following steps:

[0035] S21: obtain the peak value data and analysis data from the peak value database;

[0036] S22: analyze and judge the peak value data by sliding window, identify the median of the window and the median of the reference window in combination with the threshold value of the peak width to judge the peak value data, and the peak value data exceeding the threshold value is abnormal data, which is stored in the abnormal database.

[0037] S23: judging the normal distribution of the peak data, and storing the data not obeying the normal distribution into an abnormal database, and the normal distribution expression is as follows:

[0038]

[0039] In the above formula, u is the mean value of the peak point, and sigma is the standard deviation of the peak point data.

[0040] As preferred, the expression of the sliding window judging the peak data is as follows:

[0041]

[0042]

[0043] In the above formula, b is the reference window width, s is the identification window width, w s and w b respectively represent the window width. is the time stamp corresponding to n-b, is the time stamp corresponding to n-s, is the time stamp corresponding to n, Ths1 and Ths2 are respectively a threshold value, corresponding to the upper and lower limit range of the standard data, s is the abnormal state of the peak value, 1 is normal, -1 is abnormal, and T represents the parameter corresponding to the window.

[0044] The beneficial effect of the present application is that the signal cycle analysis technology can be used to give an early warning of the abnormal state of the unit without affecting the operation state of the unit. BRIEF DESCRIPTION OF DRAWINGS

[0045] Figure 1 The figure is a schematic diagram of an abnormal identification early warning system of a hydroelectric generating set.

[0046] Figure 2 The figure is a flow chart of an abnormal identification early warning method of a hydroelectric generating set.

[0047] Figure 3 The figure is a detailed flow chart of an abnormal identification early warning method of a hydroelectric generating set.

[0048] Figure 4 The figure is a sliding window schematic diagram of an abnormal identification early warning method of a hydroelectric generating set.

[0049] In the figure, 1 is a historical database, 2 is a peak value database, 3 is a searching module, 4 is a parameter database, 5 is a detection module, 6 is an abnormal database, 7 is an early warning module, and 8 is a processing module. DETAILED DESCRIPTION

[0050] The technical solutions of the present application are further specifically described below by examples in combination with the drawings.

[0051] An abnormality identification and early warning system of a hydroelectric generating set is shown in the following example. Figure 1 The system includes:

[0052] A historical database 1 for saving historical operation state data of the generating set; a peak value database 2 for saving data obtained by periodic analysis of the historical operation state data of the generating set; a searching module 3 for reading condition parameters for searching peak values from a parameter database 4 and then performing periodic analysis on the data in the historical database 1 to obtain peak value data and store the peak value data in the peak value database 2, connecting the historical database 1, the peak value database 2 and the parameter database 4; the parameter database 4 for storing parameters of limited conditions for searching peak values, different parameters of oil level and oil pressure peak values corresponding to different parameters of limited conditions; a detection module 5 for reading peak value data from the peak value database 2, performing real-time monitoring on the peak value data in the peak value database 2 through sliding window and timing task, storing obtained abnormal data in an abnormality database 6, connecting the peak value database 2 and the abnormality database 6; the abnormality database 6 for storing abnormal data of the detection module 5; an early warning module 7 for reading abnormal data, early warning through a monitoring page, connecting the abnormality database 6 and a processing module 8; the processing module 8 for judging abnormal data and outputting information of suggestions for solving measures.

[0053] The historical database 1 stores historical operation state data of the generating set continuously according to time, and the historical operation state data of the generating set includes real-time data such as oil pressure data of an oil tank, oil level data of an oil tank, vibration amplitude data of the generating set and vibration direction data of the generating set.

[0054] The peak value database 2 stores peak value parameter information such as peak value points, rising time and rising rate, and can analyze and judge different parameter relationships.

[0055] The searching module 3 needs to pre-process data in the whole peak value searching module to ensure accuracy of results, such as filtering and interpolation processing of data, threshold value judgment of abnormal data, etc.

[0056] The parameter database 4 stores parameters of limited conditions for searching peak values, and different parameters of oil level and oil pressure peak values correspond to different parameters of limited conditions. The main purpose of the parameter database 4 is to store parameters of limited conditions for searching peak values of different generating sets. The parameters under each condition are verified and uploaded to the parameter database 4, and each parameter of limited conditions has explanation and description, which can realize automatic processing and identification effect in cooperation with the timing task.

[0057] The detection module 5 judges the abnormality according to the data information of the search module 3 to obtain abnormal time information. The detection module 5 involves multiple steps in the entire abnormality detection module. For example, the data is analyzed by a sliding window, the sliding window size is determined, a timing task is set for the detection module, the unit is monitored in real time, and various abnormality monitoring algorithms can be selected, from peak width monitoring to normal distribution, different algorithms can judge the abnormality from different aspects and different emphases, which is directly related to the types of unit operation parameters to be solved, therefore, different abnormality analysis algorithms and different peak value parameters need to be selected to configure the peak value search module, and finally the accuracy and efficiency of the model are applied and verified;

[0058] The early warning module 7 returns the data in the abnormality database 6 to the visual window interface for warning processing.

[0059] The processing module 8 has information of common unit position parameters and solutions to common problems, such as low oil level, high unit pressure and other abnormal conditions. The abnormal state of the unit is judged by the detection module 5, the abnormal condition is returned, and the corresponding solution is output, so as to avoid delay in handling due to lack of experience or other factors.

[0060] As shown in Figure 2 and Figure 3 An abnormality identification and early warning method of a hydroelectric unit, comprising the following steps:

[0061] S1: using wavelet transform to periodically analyze and calculate the unit operation state data, and storing the peak value data into a peak value database.

[0062] S2: detecting the abnormality of the peak value data in the peak value database, and storing the abnormal data into an abnormality database;

[0063] S3: transmitting the abnormal data to a processing module, and outputting a processing scheme.

[0064] S1: using wavelet transform to periodically analyze and calculate the unit operation state data, and storing the peak value data into a peak value database.

[0065] Specifically, a noise data processing model is established to process the operation parameters of the unit; a peak value identification model is established to search for the peak value of the parameter information of the unit; a peak value data calculation model is established to analyze and process the data after searching for the peak value of the parameter information of the unit, and store the data into a peak value database.

[0066] S11: establishing a noise data processing model to perform linear interpolation and Gaussian smoothing filtering processing on the operation parameters of the unit obtained from a historical database.

[0067] Let G(x, y) represent the Gaussian convolution kernel, (i, j) is the center value of the data, and σ is the standard deviation. The representation method of the Gaussian smoothing filter is:

[0068]

[0069] In the above formula, ΔE is the difference between the (i, j) and (x, y) positions, σ is the standard deviation, and th is the threshold value. After inputting the data, first determine whether the parameter information in the neighborhood and the parameter information of the center point is greater than the threshold value. If so, set the weight value of this point to 0, otherwise, it is still the initial value. The parameter point with a weight value of 0 is interpolated, and the values of the adjacent data points are used to do mean interpolation of the parameter point.

[0070] S12: Establish a peak recognition model to find the peak value of the parameter information of the unit. Specifically, the following steps are included:

[0071] S121: The data of the unit operating state is stored in the historical database in time sequence format. Wavelet transform is used to find the peak value in the data vector, and convolution is used to smooth the data vector to filter the wave peaks with low signal-to-noise ratio, avoiding interference caused by sampling errors on the wave peak recognition.

[0072] S122: Process the data after processing, and obtain the first difference data and the subscript array with a value of 0 after the first difference. Then, the array data with subscript 0 is differentiated again to determine the level of the current waveform. The array obtained at this time is grouped and the time information is recorded. The array information of the horizontal waveform obtained is processed so that there is only one peak or trough value in the range of each horizontal waveform.

[0073] S123: Determine whether the prominence of the peak value or trough value is greater than the prominence threshold value and the width is greater than the width threshold value. If so, the peak value or trough value meets the expectation. The threshold parameters are encapsulated into a parameter interface to judge and analyze the peak value search.

[0074] In order to ensure that false peaks are not recognized in the peak recognition and judgment process, only one peak is allowed in the range, the prominence of the peak is greater than a certain threshold, and the width of the peak is greater than a certain range to limit the judgment of the peak. The peak value judgment method parameters are encapsulated into a parameter interface to judge and analyze the peak value search.

[0075] S13: After finding the peak value of the parameter information of the unit, the data is analyzed and processed to obtain the rising rate, falling rate, rising time and falling time of the periodic signal, and is sorted and stored in the peak database for subsequent abnormal peak value judgment and processing.

[0076] S2: Abnormality detection is performed on the peak data in the peak database, and the abnormal data is stored in the abnormal database.

[0077] Specifically, the peak data and analysis data of the unit operation state parameters are obtained from the peak database, the peak data is analyzed and judged through the sliding window, the accuracy and reliability of the identification result are increased, the normal distribution of the peak data is judged, and the data not subject to the normal distribution is stored in the abnormal database.

[0078] S21: The peak data and analysis data of the unit operation state parameters are obtained from the peak database;

[0079] S22: As shown in the figure, the peak data is analyzed and judged through the sliding window, the accuracy and reliability of the identification result are increased; Figure 4

[0080]

[0081]

[0082] In the formula, b is the reference window width, s is the identification window width, w s And w b Respectively represent the window width. is the time stamp corresponding to n-b, is the time stamp corresponding to n-s, is the time stamp corresponding to n. Ths1 and Ths2 are respectively a threshold value, corresponding to the upper and lower limit range of the standard data, s is the abnormal state of the peak value, 1 is normal, -1 is abnormal, and T represents the parameter corresponding to the window.

[0083] The peak width reference value of the reference window is calculated, and the peak width deviation value is used to judge whether the waveform in the identification window is abnormal. If the peak width of multiple waveforms in the identification window is abnormal, the last waveform in the window is abnormal, and the abnormal data is stored in the abnormal database;

[0084]

[0085] The median of the peak width in the identification window range.

[0086] The median of the peak width in the reference window range.

[0087] The peak data is judged by combining the median of the identification window and the median of the reference window with the threshold value of the peak width. If the peak data exceeds the threshold value, it is an abnormal waveform, and the abnormal data is stored in the abnormal database.

[0088] ​S23: judging the normal distribution of the peak data, storing the data not obeying the normal distribution into the abnormal database, and the normal distribution expression is as follows.

[0089]

[0090] In the above formula, u is the mean of the peak point, and sigma is the standard deviation of the peak point data. The obtained result is corrected by the normal distribution probability table to judge the data distribution.

[0091] S3: transmitting the abnormal data to the processing module and outputting the processing scheme.

[0092] Specifically, the warning module obtains the abnormal data of the abnormal database and performs warning, and pushes all information to the processing module, so as to facilitate the formulation of the fault processing scheme; the detection module judges the abnormal state of the unit and returns the abnormal situation, so as to output the corresponding solution, avoiding the delay of abnormal processing due to insufficient experience or other factors.

Claims

1. An abnormality recognition and early warning system for a hydroelectric generating unit, characterized in that, The application relates to a peak value detection method and device for a unit, which comprises the following parts: a historical database for storing unit historical operation state data; a peak value database for storing peak value data obtained by periodically analyzing the unit historical operation state data, wherein the peak value data comprises the rising rate, the falling rate, the rising time and the falling time of a periodic signal; a searching module for reading condition parameters for searching for peak values from a parameter database and then periodically analyzing the data in the historical database to obtain peak value data and store the peak value data in the peak value database, wherein the searching module is connected with the historical database, the peak value database and the parameter database; a parameter database for storing the limited condition parameters for searching for peak values; a detection module for reading the peak value data from the peak value database, performing real-time monitoring on the peak value data in the peak value database through a sliding window and a timing task, storing the obtained abnormal data in an abnormal database, and connecting the peak value database with the abnormal database, wherein the detection module specifically comprises the following steps: judging the peak value data by identifying the median in a window and the median in a reference window and combining the threshold of the peak width, and the peak value data exceeding the threshold is abnormal waveform; judging the peak value data to be normal distribution, and the data not obeying the normal distribution is abnormal data; an abnormal database for storing the abnormal data of the detection module; an early warning module for reading the abnormal data and performing early warning through a monitoring page, and connecting the abnormal database with a processing module; and 2. The abnormality identification and early warning system of a hydroelectric generating unit according to claim 1, characterized in that, a processing module for judging the abnormal data and outputting solution measure suggestion information.

3. The abnormality identification and early warning system of a hydroelectric generating unit according to claim 1, characterized in that, The historical database stores the unit historical operation state data in a time sequence format, and the unit historical operation state data comprises oil pressure data of an oil tank, oil level data of the oil tank, unit vibration amplitude data and unit vibration direction data.

4. The abnormality identification and early warning system of a hydroelectric generating unit according to claim 1, characterized in that, The limited condition parameters are uploaded to the parameter database after verification, the limited condition parameters match the peak value data of the unit historical operation state data, and one limited condition parameter corresponds to one or more peak value data.

5. A method for abnormality identification and early warning of a hydroelectric generating unit, applicable to the abnormality identification and early warning system of any one of claims 1-4, characterized in that, The peak value database stores the rising rate, the falling rate, the rising time and the falling time of a periodic signal. The application further discloses a peak value detection method for a unit, which comprises the following steps: S1: performing periodic analysis and calculation on unit operation state data by using wavelet transform, and storing peak value data in a peak value database; S2: performing abnormal detection on the peak value data in the peak value database, and storing abnormal data in an abnormal database; 6. The method of claim 5, wherein the method further comprises: S3: transmitting the abnormal data to a processing module, and outputting a processing scheme. The application further discloses a peak value detection method for a unit, which comprises the following steps: S11: establishing a noise data processing model, performing linear interpolation and Gaussian smoothing filter processing on the operation parameters of the unit obtained from the historical database; S12: establishing a peak value identification model, performing peak value searching on the parameter information of the unit, and obtaining peak value data; 7. The method of claim 6, wherein the method further comprises: S13: analyzing and processing the peak value data, and obtaining the parameters of the rising rate, the falling rate, the rising time and the falling time of a periodic signal and storing the parameters in the peak value database. The Gaussian smoothing filter expression is as follows: G(x,y)=th*exp(-((i-j)2+(j-k)2) / 2*sigma2) After inputting the data, it is determined whether the parameter information in the neighborhood and the parameter information of the center point are greater than the threshold value. If yes, the weight value of the point is set to 0, otherwise, it is still the initial value; The parameter point with the weight value of 0 is subjected to interpolation processing, and the values of the adjacent data points are used to do mean value interpolation of the parameter point.

8. The method of claim 6, wherein the method further comprises: A peak recognition model is established to find the peak value of the parameter information of the unit, including the following steps: S121: find the peak value in the data vector by wavelet transform, and use convolution to smooth the data vector; S122: perform transposition difference processing on the processed data to obtain the difference data and the index array with a value of 0, perform difference processing on the array again to determine the level of the waveform, group the obtained array, and record the time information, and process the array of the horizontal waveform so that there is only one peak value or trough value in the range of each horizontal waveform; S123: determine whether the prominence of the peak value or trough value is greater than the prominence threshold value and whether the width is greater than the width threshold value, if yes, the peak value or trough value meets the expectation.

9. The method of claim 5, wherein the method further comprises: Abnormal detection is performed on the peak value data in the peak value database, and the abnormal data is stored in the abnormal database, including the following steps: S21: obtain the peak value data and analysis data from the peak value database; S22: analyze and judge the peak value data by sliding window, identify the median of the window and the median of the reference window, and combine the threshold value of the peak width to judge the peak value data. If the peak value data exceeds the threshold value, it is abnormal data, which is stored in the abnormal database; S23: judge the normal distribution of the peak value data, and store the data not subject to normal distribution in the abnormal database. The normal distribution expression is as follows: In the above formula, u is the mean value of the peak point, and σ is the standard deviation of the peak point data.

10. The method of claim 9, wherein the method further comprises: The expression for judging the peak value data by sliding window is as follows: In the above formula, w s and w b respectively represent the identification window width and the reference window width, is the timestamp corresponding to n-b, is the timestamp corresponding to n-s, is the timestamp corresponding to n, Ths1 and Ths2 are respectively the lower threshold value and the upper threshold value, s is the abnormal state of the peak value, T1 and T2 respectively represent the parameters corresponding to the identification window and the reference window, 1 is normal, and -1 is abnormal.

Images