Hydroelectric generating set operation state online monitoring method and device based on big data processing

Abstract

The invention relates to the technical field of electric data processing, in particular to a hydroelectric generating set operation state online monitoring method and device based on big data processing. Determining a characteristic difference value of each modal component in each type of operation data in each working condition interval; and based on each modal component and the characteristic difference value in each type of operation data in each working condition interval, optimizing a de-trending fluctuation analysis algorithm, taking each type of operation data in a preset time period as the input of the optimized de-trending fluctuation analysis algorithm, and outputting each type of de-trending operation data so as to monitor the operation data of the hydroelectric generating set. According to the method, trend fluctuation interference caused by working condition changes is inhibited, and the accuracy of monitoring the operation data of the hydroelectric generating set is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electric data processing, in particular to a water-turbine-generator-set operation-state online monitoring method and device based on big data processing. BACKGROUND

[0002] With the decreasing of non-renewable resources, the superiority of hydroelectric power generation is increasingly displayed. As an important equipment in the water-turbine-generator-set, the water turbine plays a key role in the normal operation of the water-turbine-generator-set and is inseparable from the normal power generation work. Once the water turbine fails, it will directly affect the safe and stable operation of the power system of the hydropower station. Therefore, monitoring the operation state of the water turbine in the water-turbine-generator-set and timely debugging, repairing and maintaining can maximize the avoidance of problems in the water turbine and ensure the normal work of the hydropower station.

[0003] The water-turbine-generator-set is a typical rotating machine, and the signals such as vibration, swing and pressure have stable spectrum characteristics during normal operation. However, long-term operation can easily cause bearing wear, shaft center deviation and other faults, resulting in abnormal signal frequency amplitude. However, the signal characteristics are different under different load conditions, and the trend fluctuation caused by the change of working condition is easy to be misjudged as a fault. The traditional trend fluctuation analysis method is insufficient for complex trend fitting, and it is difficult to effectively remove the interference, which affects the accuracy of the monitoring of the operation data of the water-turbine-generator-set. SUMMARY

[0004] In order to solve the above technical problems, the purpose of the present application is to provide a water-turbine-generator-set operation-state online monitoring method and device based on big data processing, and the technical solution adopted is as follows: In the first aspect, the present application provides a water-turbine-generator-set operation-state online monitoring method based on big data processing, which comprises the following steps: Obtaining all kinds of operation data in a preset period during the operation of the water-turbine-generator-set; Respectively analyzing the average distribution and the dispersion degree of the load data and the rotating speed data in all kinds of operation data in the preset period, respectively determining the load threshold and the rotating speed threshold, respectively comparing the rotating speed data with the rotating speed threshold and the load data with the load threshold, so as to divide the preset period into multiple working condition intervals; Decomposing each kind of operation data in each working condition interval into multiple modal components, determining the impact strength of any time in each modal component based on the amplitude and the kurtosis of any time in each modal component, so as to determine the impact eigenvalue of each modal component; for each kind of operation data, based on the difference between the impact eigenvalues of each modal component in each working condition interval and the modal components in the same frequency band in the adjacent working condition interval, the characteristic difference value of each modal component in each kind of operation data in each working condition interval is determined; Based on each modal component and characteristic difference value in each type of operation data in each working condition interval, the trend-eliminating fluctuation analysis algorithm is optimized, each type of operation data in a preset period is taken as an input of the optimized trend-eliminating fluctuation analysis algorithm, and each type of trend-eliminated operation data is output, so as to monitor the operation data of the hydroelectric generating set.

[0005] Preferably, the all types of operation data include load data, speed data, vibration data, swing data, pressure and temperature of the hydroelectric generating set.

[0006] Preferably, the determination method of the load threshold value and the speed threshold value is: Load data and speed data of the hydroelectric generating set at all times within a preset length before a preset period are acquired, and the mean value and the standard deviation of the load data at all times and the mean value and the standard deviation of the speed data at all times are respectively calculated. The sum of the mean value and the standard deviation of the load data and the sum of the mean value and the standard deviation of the speed data are respectively denoted as a load threshold value and a speed threshold value.

[0007] Preferably, the preset period is divided into multiple working condition intervals, including: If the load data at time i is greater than the load threshold value and the speed data is greater than the speed threshold value within the preset period, the time i is recorded as a boundary time, all boundary times within the preset period are obtained by traversing all times within the preset period, and a time period between two adjacent boundary times is taken as a working condition interval.

[0008] Preferably, the impact intensity of any time in each modal component is a result of positive fusion of the amplitude and the kurtosis of any time in each modal component.

[0009] Preferably, the impact characteristic value of each modal component is a mean value of the impact intensity of all frequencies in each modal component.

[0010] Preferably, the expression of the characteristic difference value of each modal component in each type of operation data in each working condition interval is: ; in the formula, represents the characteristic difference value of the modal component h in the gth type of operation data in the kth working condition interval; 、 represents the impact characteristic value of the modal component h in the gth type of operation data in the kth working condition interval; represents the impact characteristic value of the modal component in the same frequency band as the modal component h in the gth type of operation data in the k-1th working condition interval; and exp( ) represents an exponential function with a natural constant as a base.

[0011] Preferably, the optimized trend-eliminating fluctuation analysis algorithm includes: The fluctuation sequence of each modal component in each type of operation data in each operating condition interval is obtained, the fluctuation sequence is equally divided into a plurality of subsequences, the fitting function of each subsequence is obtained by fitting the subsequence respectively, the fluctuation function in the detrended fluctuation analysis algorithm is optimized based on the difference between each subsequence and its fitting function and in combination with the characteristic difference value, and specifically: The expression of the fluctuation function optimized based on the gth type of operation data is: ; in the formula, h k g represents the modal component h in the gth type of operation data in the kth operating condition interval; h k g represents the characteristic difference value of the modal component h in the gth type of operation data in the kth operating condition interval; h k g represents the fitting function of the nth subsequence of the modal component h in the gth type of operation data in the kth operating condition interval; h k g represents the number of all subsequences of the modal component h in the gth type of operation data in the kth operating condition interval; h k g represents the number of all modal components in the gth type of operation data in the kth operating condition interval; h represents the number of all operating condition intervals in the preset time period.

[0012] Preferably, the monitoring of the operation data of the hydroelectric generating set comprises: All types of operation data in a preset time period after the hydroelectric generating set is put into operation for the first time are obtained, and all types of detrended operation data in the preset time period are obtained according to the obtaining mode of each type of detrended operation data in the preset time period; Reference data of each type of detrended operation data in each operating condition interval in the preset time period in the corresponding type of detrended operation data in the preset time period is obtained, the mean and variance of all reference data are calculated, the result of adding 3 times the variance to the mean is recorded as a reference threshold value, and each type of detrended operation data in each operating condition interval is taken as the input of a sliding window algorithm. If the mean of each type of detrended data in the sliding window is greater than the corresponding reference threshold value, the corresponding type of detrended operation data of the hydroelectric generating set in the sliding window is abnormal, otherwise, the corresponding type of detrended operation data of the hydroelectric generating set in the sliding window is normal.

[0013] In a second aspect, the embodiments of the present application also provide a hydroelectric generating set operation state online monitoring device based on big data processing, which comprises a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, the steps of the hydroelectric generating set operation state online monitoring method based on big data processing are implemented.

[0014] The present application has at least the following beneficial effects: ​The application sets a threshold by analyzing the mean and standard deviation of load and speed data, identifies the demarcation time of working condition change, thereby divides the preset time period into multiple relatively stable working condition intervals, which helps to improve the pertinence and accuracy of subsequent data analysis; further, the application decomposes the running data into multiple modal components, calculates the impact strength and characteristic difference value in combination with the amplitude and kurtosis, effectively improves the insufficient fitting problem of traditional detrended fluctuation analysis under complex working conditions, realizes the accurate characterization of signal characteristics of hydroelectric generating units under different working conditions, significantly improves the distinguishing ability of normal physical response and potential failure, and enhances the accuracy of state monitoring; further, the application introduces the characteristic difference value to optimize the detrended fluctuation analysis algorithm, effectively suppresses the trend fluctuation interference caused by working condition change, improves the accuracy of signal detrending processing, and combines the sliding window and 3 The principle establishes a dynamic threshold monitoring mechanism, realizes real-time and accurate monitoring of the running state of the hydroelectric generating unit, and improves the accuracy of monitoring the running data of the hydroelectric generating unit. BRIEF DESCRIPTION OF DRAWINGS

[0015] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the application or the prior art, the drawings needed to be used in the embodiments or the prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the application, and for those skilled in the art, other drawings can be obtained without creative labor on the basis of these drawings.

[0016] Figure 1 The step flow chart of the online monitoring method of the running state of the hydroelectric generating unit based on big data processing provided by one embodiment of the application is shown in the figure. Figure 2 The characteristic difference value acquisition process flow chart provided by one embodiment of the application is shown in the figure. DETAILED DESCRIPTION

[0017] In order to further illustrate the technical means and effects adopted by the application to achieve the predetermined invention purpose, the online monitoring method and device of the running state of the hydroelectric generating unit based on big data processing according to the application are described in detail below in combination with the drawings and preferred embodiments. The specific implementation, structure, features and effects of the online monitoring method and device of the running state of the hydroelectric generating unit based on big data processing according to the application are described in detail below. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0018] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the application belongs.

[0019] The application provides a big data processing-based online monitoring method and device for the running state of a hydroelectric generating set.

[0020] Please refer to Figure 1 which shows a step flowchart of a big data processing-based online monitoring method for the running state of a hydroelectric generating set, and the method comprises the following steps. Step S1: acquiring all kinds of running data in a preset time period during the running of the hydroelectric generating set.

[0021] Various sensors are arranged on the equipment of the hydroelectric generating set to collect all kinds of running data in a preset time period during the running of the hydroelectric generating set, wherein the all kinds of running data include load data, speed data, vibration data, swing data, pressure and tile temperature of the hydroelectric generating set. In this embodiment, the load data is acquired at the active power measuring point of the generator outlet of the hydroelectric generating set, the speed data of the hydroelectric generating set is collected by using a key phase sensor, the vibration data during the running of the hydroelectric generating set is acquired by using a vibration sensor, the swing data during the running of the hydroelectric generating set is collected by using an eddy current sensor, the pressure sensor of the hydroelectric generating set is collected by using a pressure sensor, and the tile temperature during the running of the hydroelectric generating set is acquired by using a temperature sensor. The data collection frequency of the above-mentioned pressure sensor, key phase sensor, eddy current sensor and temperature sensor is f1, the collection frequency of the vibration data is f2, and the collection frequency of the load data is f3. The data collection of the above-mentioned sensors starts synchronously.

[0022] It should be noted that the length of the preset time period and the value of the data collection frequency f are artificially set. In this embodiment, the length of the preset time period is 10 min, the value of the data collection frequency f1 is 500 Hz, the value of the collection frequency f2 is 20 kHz, and the value of the collection frequency f3 is 50 Hz. In actual application, as other implementation manners, the implementer can also set them by himself according to the specific situation, and this embodiment does not have special limitations.

[0023] Step S2: dividing the preset time period into different working condition intervals according to the load and speed data, constructing a characteristic value difference by analyzing the impact characteristics of the data modal components in each working condition interval and the difference between adjacent working conditions.

[0024] As a peak shaving power station, the water turbine unit of the hydropower station often adjusts the working condition according to the change of the power load to meet the demand of the power grid. Under different load conditions, the rotating speed of the water turbine in the water turbine unit is different, and the frequency and amplitude of the detected vibration, swing, pressure and other data are also different. With the change of the working condition, these data signals show a trend change. For example, when the load increases, the resistance torque of the water turbine increases, the rotating speed decreases, the force on the runner is uneven, the hydraulic impact significantly increases, the vibration frequency of the vibration data increases, the vibration amplitude also shows a trend increase due to the mechanical stress of the guide vane, and the pressure data also increases. Under the action of various factors, with the change of the working condition of the water turbine unit, the signal also shows a corresponding trend change. However, this trend change caused by the change of the load condition will be misdetected as an abnormal fault signal. In order to accurately monitor the running state of the water turbine unit, the data signal needs to be de-trended. However, the trend change is complex, which not only includes the trend change of the water turbine operation, but also includes the trend change of the hydraulic impact caused by the trend of the water turbine. The traditional de-trended fluctuation analysis algorithm (DFA) has insufficient fitting ability for the trend fluctuation under the complex frequency structure, resulting in insufficient removal of the trend items of the signal.

[0025] Therefore, in order to solve the above problems, the present application divides different working condition intervals, constructs the feature value difference of the IMF component of the same frequency under different working conditions to represent the distribution and fluctuation degree of each type of running data under different working conditions, and constructs the feature value difference value. The specific process is as follows: S201: respectively analyze the average distribution and dispersion degree of the load data and the rotating speed data in all types of running data in a preset period, respectively determine the load threshold and the rotating speed threshold, respectively compare the rotating speed data with the rotating speed threshold and the load data with the load threshold, and divide the preset period into multiple working condition intervals.

[0026] Considering the load data and the rotating speed data of the water turbine unit under different working conditions, the working condition of the water turbine data collected at all times is segmented and divided. In this embodiment, by respectively analyzing the average distribution and dispersion degree of the load data and the rotating speed data in all types of running data in a preset period, respectively determining the load threshold and the rotating speed threshold, respectively comparing the rotating speed data with the rotating speed threshold and the load data with the load threshold, and dividing the preset period into multiple working condition intervals, each segment of data corresponds to a relatively stable working condition interval. The specific division process of the working condition interval is as follows: Firstly, the load data and the rotating speed data of the water turbine unit at all times within a preset time period before the preset period are obtained, and the mean and standard deviation of the load data at all times and the mean and standard deviation of the rotating speed data at all times are calculated. It should be noted that the value of the preset time length is artificially set, and in the embodiment, the value of the preset time length is 7 days. In actual application, as another implementation manner, the implementer can set it by himself according to the specific situation, and the embodiment does not have special limitation.

[0027] Further, the sum of the mean and the standard deviation of the load data and the sum of the mean and the standard deviation of the speed data are respectively denoted as a load threshold and a speed threshold. Further, if the load data at time i is greater than the load threshold and the speed data is greater than the speed threshold within the preset time period, time i is recorded as a boundary time, all boundary times within the preset time period are obtained by traversing all times within the preset time period, and a time period between adjacent two boundary times is taken as a working condition interval.

[0028] By now, the embodiment sets the threshold by analyzing the mean and the standard deviation of the load and speed data, identifies the boundary time of the working condition change, and thus divides the preset time period into multiple relatively stable working condition intervals, which is helpful to improve the pertinence and accuracy of subsequent data analysis.

[0029] S202: decompose each type of running data in each working condition interval into multiple modal components, determine the impact intensity at any time in each modal component based on the amplitude and kurtosis at the time, and determine the impact feature value of each modal component; for each type of running data, determine the feature difference value of each modal component in each type of running data in each working condition interval based on the difference between the impact feature values of each modal component in each working condition interval and the modal components in the same frequency band in the adjacent working condition interval.

[0030] Further, trend analysis is performed based on the data segmented by the above working condition intervals: under the normal load working condition of the hydroelectric generating set, the water turbine is in a stable working state, and multiple types of running data, such as vibration, swing, pressure, tile temperature, etc., have small fluctuations, concentrated frequency spectrum, and insignificant trend changes. The vibration data mainly shows the response of the structural natural frequency, and the overall data signal features present a stable state; under the power peak shaving demand, the hydroelectric generating set enters the working condition state of variable load, which causes the trend change of the load data; under the low load working condition, the water turbine has low water energy absorption efficiency, and part of the energy that is not effectively converted is released in the form of irregular mechanical vibration, which causes the increase of vibration amplitude, the enhancement of pressure fluctuation, and the rise of swing; under the high load working condition, the increase of power causes the increase of water turbine resistance torque and the decrease of rotating speed, the uneven force on the runner, and the significant enhancement of water impact, which causes the rise of vibration frequency and pressure pulse. In order to meet the demand of the output power of the generator, the rotating frequency of the hydroelectric generator is adjusted by the speed regulator, and at the same time, the heat generation of the internal bearing of the mechanical rotation is increased, which causes the gradual rise of tile temperature.

[0031] The trend changes of the vibration, swing, and pressure data are not only affected by the rotation speed of the hydraulic turbine, but also introduced by the trend changes of the hydraulic impact into more complex trends. The traditional detrend fluctuation analysis algorithm performs trend fluctuation fitting based on piecewise linear regression. However, the trend changes of the signal during the operation of the hydroelectric generating set are disturbed by very complex nonlinear working conditions. The vibration data, swing data, and pressure data themselves show a slow rising form. The influence of the hydraulic impact also increases with the change of the working condition, and the influence of the hydraulic impact has obvious multi-frequency components. The influence of the hydraulic impact is a low-frequency signal, and the change of the vibration caused by the influence of the hydraulic impact is a high-frequency signal. Therefore, there are many trend change signals in the low-frequency band, and the vibration changes caused by the abnormality are mainly in the form of high-frequency impact. The detrend fluctuation analysis algorithm performs linear fitting of the trend fluctuation function on the time series signal in the time domain, and the fitting capability of the trend fluctuation under the complex frequency structure is insufficient, which leads to insufficient detrending in the complex trend changes.

[0032] Therefore, in the embodiment, each type of operation data in each working condition interval is decomposed into a plurality of modal components. The impact strength of any time point in each modal component is determined based on the amplitude and kurtosis of the time point in the modal component, so as to determine the impact characteristic value of each modal component. For each type of operation data, the characteristic difference value of each modal component in each type of operation data in each working condition interval is determined based on the difference between the impact characteristic values of each modal component in each working condition interval and the modal components in the same frequency band in the adjacent working condition interval, so as to improve the traditional detrend analysis, and more accurately identify the real operation state of the hydroelectric generating set. Specifically, In the embodiment, first, each type of operation data in each working condition interval is taken as the input of the modal decomposition algorithm, and all modal components are output. In the embodiment, the modal decomposition algorithm used is variational modal decomposition. In actual application, the implementer can also use other modal decomposition methods such as empirical modal decomposition algorithm according to the specific situation. The selection of the modal decomposition algorithm is not specially limited in the embodiment.

[0033] The variational modal decomposition algorithm is a known technology, and the specific process of dividing the operation data into a plurality of modal components will not be described here.

[0034] Further, the impact strength of any time point in each modal component is determined based on the amplitude and kurtosis of the time point in the modal component, which is used to represent the comprehensive representation of the energy and impact of the modal component, and reflects the contribution strength of the modal component to the overall behavior of the signal at a certain time point. Specifically, In the embodiment, the result of forward fusion of the amplitude and kurtosis at any moment in each modal component is taken as the impact strength at any moment in each modal component. If the impact strength of the current class running data current modal component is greater, it indicates that the energy of the current class running data at the frequency of the modal component is strong and changes dramatically, and there may be abnormal states such as mechanical impact, bearing failure, water impact or structural resonance. Conversely, if the impact strength of the current class running data current modal component is smaller, it indicates that the energy of the current class running data at the frequency of the modal component is weak and changes smoothly, and the signal in this frequency band shows stable, periodic or random noise characteristics, which usually reflects that the device is in a normal operating state or that the frequency component is not significantly disturbed, and there is no obvious failure sign or abnormal impact event.

[0035] The calculation method of kurtosis is a known technology, and the specific calculation process will not be described again.

[0036] It should be understood that forward fusion refers to combining two or more indicators together through addition or multiplication or the like in order to obtain a comprehensive indicator, so as to more comprehensively and accurately evaluate a phenomenon or a problem. This fusion method is not limited to simple arithmetic operations, but can also include more complex statistical models and analysis methods, and the implementer can select them according to the specific circumstances, and the embodiment does not make special limitations.

[0037] Preferably, the product of the amplitude and kurtosis at any moment in each modal component is taken as the impact strength at any moment in each modal component in the embodiment. In actual application, as other implementation manners, the implementer can also use other forward fusion methods such as sum value according to the specific circumstances, and the embodiment does not make special limitations.

[0038] Further, the impact characteristic value of each modal component is determined based on the impact strength at any moment in each modal component, and specifically: In the embodiment, the mean value of the impact strength of all frequencies in each modal component is taken as the impact characteristic value of each modal component. The impact characteristic value is used to represent the average behavior characteristics of the modal component in the working condition interval, and is a general evaluation of the activity degree or disturbance level of the modal component in the whole time period. If the impact characteristic value of the current modal component is greater, it indicates that the current modal component is continuously active in the whole working condition interval, and the energy and impact are strong, and there may be continuous mechanical failure, bearing wear or water impact and other abnormalities. Conversely, if the impact characteristic value of the current modal component is smaller, it indicates that the current modal component is relatively stable in the whole working condition interval, and the energy and impact are weak, and the signal as a whole shows low activity or low disturbance level, which usually indicates that there is no continuous or significant abnormal event in the frequency band, and the device is in a stable operating state in the working condition interval, and there is no major failure risk or long-term deterioration trend.

[0039] Further, the embodiment is directed to each type of operating data, based on the difference between the impact characteristic value of each modal component under each working condition interval and the impact characteristic value of the modal component in the same frequency band under the adjacent working condition interval, to determine the characteristic difference value of each modal component in each type of operating data in each working condition interval, specifically: As an implementation, in the embodiment, the expression of the characteristic difference value of the modal component h in the gth type of operating data in the kth working condition interval is: ; in the formula, represents the characteristic difference value of the modal component h in the gth type of operating data in the kth working condition interval; , represents the impact characteristic value of the modal component h in the gth type of operating data in the kth working condition interval; represents the impact characteristic value of the modal component in the same frequency band as the modal component h in the gth type of operating data in the k-1th working condition interval; exp() represents the exponential function with the natural constant as the base.

[0040] Preferably, the process flow chart of the characteristic difference value acquisition process provided by the embodiment is as shown in Figure 2 .

[0041] According to the characteristic difference value of each modal component in each type of operating data in each working condition interval, it can be understood that the characteristic value difference reflects the influence degree of working condition change on signal characteristics. If the difference between the characteristic difference value of the modal component h in the gth type of operating data in the kth working condition interval and the impact characteristic value of the modal component in the same frequency band as the modal component h in the gth type of operating data in the k-1th working condition interval is smaller, it means that the gth type of operating data in the kth working condition interval changes more smoothly, and therefore, the corresponding characteristic value difference is smaller, which means that the signal characteristics between adjacent working condition intervals are close, the working condition changes gently, and the equipment runs stably. On the contrary, if the characteristic difference value is larger, it means that the signal characteristics in the frequency band of the modal component have changed significantly in the kth working condition interval relative to the k-1th working condition interval, which may be a normal physical response such as working condition driven hydraulic impact or mechanical stress mutation, or a potential fault induced by working condition change, which should be paid attention to and further analyzed.

[0042] So far, by decomposing the operating data into multiple modal components, and combining the impact strength and characteristic difference value calculated by amplitude and kurtosis, the embodiment effectively improves the fitting problem of traditional detrended fluctuation analysis under complex working conditions, realizes the accurate description of signal characteristics of hydroelectric generating units under different working conditions, significantly improves the differentiation ability of normal physical response and potential fault, and enhances the accuracy of state monitoring.

[0043] Step S3: Based on each modal component and the characteristic difference value in each type of operating data in each operating condition interval, the detrended fluctuation analysis algorithm is optimized, each type of operating data in a preset time period is taken as the input of the optimized detrended fluctuation analysis algorithm, and each type of detrended operating data is output, so as to monitor the operating data of the hydroelectric generating set.

[0044] According to the characteristic difference value obtained in step S2, the fluctuation sequence of each modal component in each type of operating data in each operating condition interval is obtained, the fluctuation sequence is equally divided into a plurality of subsequences, each subsequence is fitted to obtain a fitting function, the fluctuation function in the detrended fluctuation analysis algorithm is optimized based on the difference between each subsequence and its fitting function and in combination with the characteristic difference value, and the specific optimization process is as follows: Firstly, in the embodiment, the result of subtracting the mean value of all amplitudes from all amplitudes of each modal component in each type of operating data in each operating condition interval is arranged in time sequence to form the fluctuation sequence of each modal component, and further, the fluctuation sequence of each modal component is divided into S equal-length subsequences. Each subsequence of each modal component in each type of operating data in each operating condition interval is fitted to obtain a subsequence, each subsequence is fitted to obtain a fitting function, the fluctuation function in the detrended fluctuation analysis algorithm is optimized based on the difference between each subsequence and its fitting curve and in combination with the characteristic difference value, and the specific optimization process is as follows: The expression of the fluctuation function of the gth type of operating data optimized based on the gth type of operating data is as follows: ; in the formula, indicates the modal component h in the gth type of operating data in the kth operating condition interval; indicates the characteristic difference value of the modal component h in the gth type of operating data in the kth operating condition interval; indicates the fitting function of the nth subsequence of the modal component h in the gth type of operating data in the kth operating condition interval; indicates the number of all subsequences of the modal component h in the gth type of operating data in the kth operating condition interval; indicates the number of all modal components of the gth type of operating data in the kth operating condition interval; indicates the number of all operating condition intervals in a preset time period.

[0045] Further, the value of S is changed repeatedly, and the calculation steps of the fluctuation function are repeated, wherein, ​In this embodiment, the value of d is 20. In actual application, as other implementation methods, the implementer can also set it according to the specific situation. The initial value of S is 10, and it is added to the previous value by 10. For example, the initial value of S is 10, and the next value after the initial value is 20. This is accumulated until S=200, and the training of the detrended fluctuation analysis algorithm is completed, and the final optimized detrended fluctuation analysis algorithm is obtained.

[0046] It should be understood that the characteristic difference value It can correct the impact of trend changes caused by changes in operating conditions, when 1. Correction of the first The complex fluctuation trend of the operating condition signal is caused by changes in the operating condition. The greater the difference in the characteristic values ​​of the IMF component between two consecutive operating conditions, the better. The larger the value, the more significant the enhancement. Here, the enhancement means not only considering the trend changes of the turbine's own equipment signals, but also the trend changes brought about by external hydraulic shocks. Enhancing the fitting of fluctuation trends can reduce the impact of external hydraulic shocks caused by changes in operating conditions.

[0047] Furthermore, in this embodiment, each type of operational data within a preset time period is used as input to the optimized detrended fluctuation analysis algorithm, and each type of detrended operational data is output. Based on the detrended operational data, the operational data of the hydropower unit is monitored. Specifically: In this embodiment, all types of operational data within a preset period after the hydropower unit is first put into operation are obtained from the system database, and all types of detrended operational data within the preset period are obtained according to the method of obtaining each type of detrended operational data within the preset period. Obtain reference data for each type of detrended operational data in each operating condition interval within a preset time period, and calculate the mean and variance of all reference data. The principle is to take the mean plus three times the variance as the reference threshold, and use the detrended data of each type in each working condition interval as the input of the sliding window algorithm. If the mean of each type of detrended data in the sliding window is greater than the corresponding reference threshold, then the detrended data of the corresponding type in the sliding window is abnormal; otherwise, the detrended data of the corresponding type in the sliding window is normal.

[0048] In this embodiment, the size of the sliding window is set to 5 seconds. In actual applications, as other implementation methods, the implementer can set the size according to the specific situation. This embodiment does not impose any special restrictions. In addition, 3 The principle is a well-known technology, and its specific principles will not be elaborated here.

[0049] It is to be supplemented that the specific process of obtaining the reference data of each type of de-trend operation trend data in each working condition interval in the preset period to the corresponding de-trend operation data in the preset period is as follows: The judgment standard of the reference data is: in the preset period, if there are data of the same continuous period in the de-trend load data and the de-trend speed data, the length of the continuous data is the same as the length of a working condition interval in the preset period, and after being arranged in time sequence, the absolute difference between the de-trend load data in the continuous period and the de-trend load data in the corresponding working condition interval is less than the load threshold in step S2, and the absolute difference between the de-trend speed data in the continuous period and the de-trend speed data in the corresponding working condition interval is less than the speed threshold in step S2, then each type of de-trend operation data in the continuous period is taken as the reference data of the corresponding type of de-trend operation data in the working condition interval.

[0050] Up to now, the embodiment introduces the feature difference value to optimize the de-trend fluctuation analysis algorithm, effectively suppresses the trend fluctuation interference caused by the working condition change, improves the accuracy of the signal de-trend processing, and combines the sliding window with the 3 The principle establishes a dynamic threshold monitoring mechanism, realizes real-time, accurate and robust monitoring of the operation state of the hydroelectric generating set, and significantly improves the accuracy of monitoring the operation data of the hydroelectric generating set.

[0051] Based on the same inventive concept as the above method, the embodiments of the present application also provide a water and electricity unit operation state online monitoring device based on big data processing, which comprises a memory, a processor and a computer program stored in the memory and running on the processor, and the processor implements the steps of any one of the above water and electricity unit operation state online monitoring methods based on big data processing.

[0052] It should be noted that the above-mentioned embodiments of the present application are only for description, and do not represent the advantages and disadvantages of the embodiments. Moreover, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are also possible or may be advantageous.

[0053] Each embodiment in the specification is described in a progressive manner, and the same or similar parts between each embodiment can be referred to each other. Each embodiment focuses on the difference from other embodiments.

[0054] The above is only the preferred embodiment of the present application, and does not limit the present application. Any modification, equivalent replacement, improvement, etc. made within the principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for online monitoring of the operating status of hydropower units based on big data processing, characterized in that, The method includes the following steps: Acquire all types of operational data within a preset time period during the operation of the hydropower unit; The average distribution and dispersion of load data and speed data in all types of operating data within the preset time period are analyzed respectively. The load threshold and speed threshold are determined respectively. The speed data and speed threshold, as well as the load data and load threshold, are compared respectively to divide the preset time period into multiple operating condition intervals. Each type of operating data under each operating condition interval is decomposed into multiple modal components. Based on the amplitude and kurtosis of each modal component at any time, the impact intensity of each modal component at any time is determined to determine the impact characteristic value of each modal component. For each type of operating data, based on the difference in impact characteristic values ​​between each modal component under each operating condition interval and the modal components in the same frequency band under adjacent operating condition intervals, the characteristic difference value of each modal component in each type of operating data under each operating condition interval is determined. Based on the modal component and characteristic difference value of each type of operating data in each operating condition interval, the detrended fluctuation analysis algorithm is optimized. Each type of operating data in the preset time period is used as the input of the optimized detrended fluctuation analysis algorithm, and each type of detrended operating data is output to monitor the operating data of the hydropower unit.

2. The online monitoring method for the operating status of hydropower units based on big data processing as described in claim 1, characterized in that, All types of operating data include: load data, speed data, vibration data, swing data, pressure, and bearing temperature of the hydropower unit.

3. The online monitoring method for the operating status of hydropower units based on big data processing as described in claim 1, characterized in that, The method for determining the load threshold and the speed threshold is as follows: Obtain the load and speed data of the hydropower unit at all times within a preset time period before the preset time period, and calculate the mean and standard deviation of the load data at all times, as well as the mean and standard deviation of the speed data at all times. The sum of the mean and standard deviation of the load data, and the sum of the mean and standard deviation of the speed data, are respectively denoted as the load threshold and the speed threshold.

4. The online monitoring method for the operating status of hydropower units based on big data processing as described in claim 1, characterized in that, The process of dividing the preset time period into multiple working condition intervals includes: If, within a preset time period, the load data at time i is greater than the load threshold and the speed data is greater than the speed threshold, then time i is recorded as the boundary time. All boundary times are obtained by iterating through all times within the preset time period, and the time period between two adjacent boundary times is taken as the operating condition interval.

5. The online monitoring method for the operating status of hydropower units based on big data processing as described in claim 1, characterized in that, The impact intensity at any time in each modal component is the result of a positive fusion of the amplitude and kurtosis at any time in each modal component.

6. The online monitoring method for the operating status of hydropower units based on big data processing as described in claim 1, characterized in that, The impact characteristic value of each modal component is the average impact intensity at all frequencies in each modal component.

7. The online monitoring method for the operating status of hydropower units based on big data processing as described in claim 1, characterized in that, The expression for the characteristic difference value of each modal component in each type of operating data within each operating condition interval is: In the formula, This represents the characteristic difference value of modal component h in the g-th type of operating data within the k-th operating condition interval; , This represents the impact characteristic value of the modal component h in the g-th type of operating data within the k-th operating condition interval; represents the impact characteristic value of the modal component in the same frequency band as the modal component h in the g-th type of operating data within the k-1th operating condition interval; exp() represents the exponential function with the natural constant as the base.

8. The online monitoring method for the operating status of hydropower units based on big data processing as described in claim 1, characterized in that, The optimized detrending volatility analysis algorithm includes: The fluctuation sequence of each modal component in each type of operating data within each operating condition interval is obtained. The fluctuation sequence is divided into multiple subsequences, and a fitting function is obtained by fitting each subsequence. Based on the difference between each subsequence and its fitting function, and combined with the feature difference value, the fluctuation function in the detrended fluctuation analysis algorithm is optimized, specifically as follows: The fluctuation function optimized based on the g-th class of running data The expression is: In the formula, This represents the modal component h in the g-th type of operating data within the k-th operating condition interval; This represents the characteristic difference value of modal component h in the g-th type of operating data within the k-th operating condition interval; The fitting function represents the nth subsequence of modal component h in the g-th type of operating data within the k-th operating condition interval; This represents the number of all subsequences of modal component h in the g-th type of operating data within the k-th operating condition interval; This represents the number of all modal components of the g-th type of operating data within the k-th operating condition interval; This indicates the number of all operating condition intervals within a preset time period.

9. The online monitoring method for the operating status of hydropower units based on big data processing as described in claim 1, characterized in that, The monitoring of the operating data of the hydropower unit includes: Obtain all types of operational data within a preset period after the hydropower unit is put into operation for the first time. Obtain all types of detrended operational data within the preset period according to the method of obtaining each type of detrended operational data within the preset period. Obtain reference data for each type of detrended operating data in each operating condition interval within a preset time period, and calculate the mean and variance of all reference data. Add three times the variance to the mean and record it as the reference threshold. Use each type of detrended operating data in each operating condition interval as input to the sliding window algorithm. If the mean of each type of detrended operating data in the sliding window is greater than the corresponding reference threshold, the corresponding type of detrended operating data of the hydropower unit in the sliding window is abnormal; otherwise, the corresponding type of detrended operating data of the hydropower unit in the sliding window is normal.

10. An online monitoring device for the operating status of a hydropower unit based on big data processing, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the online monitoring method for the operating status of hydropower units based on big data processing as described in any one of claims 1-9.

Images