Pumped storage unit degradation trend integrated prediction method and system

By combining time domain and frequency domain health models, using multiple working condition parameters and vibration data, a comprehensive deterioration trend sequence is constructed, and using Autoformer and ARMA models to predict, the insufficient selection of working condition parameters and time scales in the existing technology are solved. The problem of neglecting differences has achieved higher accuracy prediction of the deterioration trend of pumped storage units.

CN118916630APending Publication Date: 2024-11-08ELECTRIC POWER RESEARCH INSTITUTE OF STATE GRID SHANDONG ELECTRIC POWER COMPANY
View PDF 9 Cites 0 Cited by

Patent Information

Application Number
CN202410929942.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-11
Publication Date
2024-11-08

AI Technical Summary

Technical Problem

In the existing pumped storage unit deterioration trend prediction methods, fewer working conditions are selected, and the constructed deterioration indicators are random, and the differences in different time scales of components are ignored, resulting in low prediction accuracy and cannot fully reflect the actual operation.

Method used

The time-domain health model and frequency-domain health model are combined to construct a comprehensive degradation trend sequence through time-sequence convolution network and ensemble empirical modal decomposition, approximate entropy and K-mean clustering algorithms, and the nonlinear and linear components are predicted by Autoformer and ARMA models respectively. , perform weighted superposition to obtain the final prediction result.

Benefits of technology

A more accurate and reasonable prediction of the deterioration trend of pumped storage units is achieved, the prediction accuracy and stability are improved, and the unit's deterioration information is fully reflected.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118916630A_ABST
    Figure CN118916630A_ABST
Patent Text Reader

Abstract

The invention discloses a pumped storage unit degradation trend integrated prediction method and system, and the method comprises the steps: obtaining historical operation data of different working conditions and corresponding historical state data, and constructing a time domain health model; performing modal decomposition on the historical state data, obtaining a frequency domain feature vector according to the approximate entropy of each modal component, clustering the frequency domain feature vector to obtain a health center vector, and constructing a frequency domain health model; a time domain health model and a frequency domain health model are adopted to obtain a time domain degradation index and a frequency domain degradation index respectively, and a comprehensive degradation trend sequence is obtained based on the time domain degradation index and the frequency domain degradation index; and decomposing the comprehensive degradation trend sequence, respectively predicting a nonlinear component and a linear component, and superposing prediction results to obtain a final degradation trend prediction result. The method comprehensively considers the change characteristics and trend of the operation state of the unit in the time-frequency domain, comprehensively reflects the degradation information of the unit, and more reasonably and accurately predicts the degradation trend of the pumped storage unit.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of health management of pumped storage units, and in particular to a method and system for integrated prediction of degradation trends of pumped storage units. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] Pumped storage units are key equipment for hydropower energy conversion. As the cumulative operating time increases, the units are prone to abnormal vibration, equipment exhaustion, unit degradation, etc. As the degree of degradation continues to increase, the equipment performance of the unit will gradually decline, which may eventually lead to equipment failure. As the online monitoring system of the pumped storage unit status collects more and more data, effectively mining the relationship between monitoring data and unit status and accurately predicting the degradation trend is a new research direction.

[0004] Patent CN114880917A proposes a method for constructing a health status model of a pumped storage unit and predicting performance trends, using active power, working head and guide vane opening as operating parameters, unit vibration as performance parameters, and establishing a standard health model of the unit through Gaussian regression, and then predicting the relative deviation between the online measured vibration and the predicted healthy vibration. However, this method only considers unit vibration and its three influencing factors: active power, head and guide vane opening, and the standard health model constructed by this method is far from comprehensive.

[0005] Patent CN117435988A proposes a vibration degradation prediction method for hydropower units in energy storage power stations based on EEMD-GRU-Transformer, which performs modal decomposition on the degradation time series and decomposes it into stable components. The components are input into the GRU-Transformer network to obtain the prediction trend of each component, and the components are superimposed to obtain the final degradation trend. However, this method uses the same model for prediction superposition, ignoring the differences in different time scales of the decomposed components, resulting in the need for further improvement in prediction accuracy.

[0006] In summary, the current methods for predicting the degradation trend of units have problems such as too few selected operating parameters, great randomness in the constructed degradation indicators, ignoring the differences in different time scales of components resulting in low prediction accuracy, and being unable to fully reflect the actual operation of pumped-storage units. Summary of the invention

[0007] In order to solve the above problems, the present invention proposes a method and system for integrated prediction of degradation trend of pumped storage units, which combines the time domain health model and the frequency health model to construct a comprehensive degradation trend sequence, comprehensively considers the changing characteristics and trends of the unit's operating status in the time and frequency domains, can comprehensively reflect the degradation information of the unit, and can more reasonably and accurately predict the degradation trend of the pumped storage unit.

[0008] In order to achieve the above object, the present invention adopts the following technical solution:

[0009] In a first aspect, the present invention provides a method for integrated prediction of degradation trend of a pumped storage unit, comprising:

[0010] Obtain the historical operation data and corresponding historical status data of the pumped storage unit under different working conditions, and use them to build a time domain health model;

[0011] Perform modal decomposition on historical state data, obtain the frequency domain feature vector under the healthy state according to the approximate entropy of each modal component, cluster the frequency domain feature vector to obtain the health center vector, and thus construct the frequency domain health model;

[0012] Obtain real-time operation data, use time domain health model and frequency domain health model to obtain time domain degradation index and frequency domain degradation index respectively, and obtain comprehensive degradation trend sequence based on the two types of degradation indexes;

[0013] The comprehensive degradation trend sequence is decomposed, the obtained nonlinear component and linear component are predicted respectively, and the prediction results are superimposed to obtain the final degradation trend prediction result.

[0014] As an optional implementation, the operating data includes water head, active power, reactive power, excitation voltage, excitation current and guide vane opening; the status data includes vibration data; a time domain health model is constructed by using a time series convolutional network with historical operating data as input and historical status data as output.

[0015] As an optional implementation, the process of obtaining the time domain degradation index using the time domain health model includes:

[0016] The real-time operation data is input into the time domain health model to obtain the health value Y(t) under the current working condition, and the relative error between the actual value T(t) and the health value Y(t) is taken as the time domain degradation index R 1 (t):

[0017] As an optional implementation method, the process of obtaining the frequency domain degradation index using the frequency domain health model includes:

[0018] The frequency domain feature vector L(t) is extracted from the status data corresponding to the real-time operation data, and the Euclidean distance between L(t) and the health center vector Ω is calculated to obtain the frequency domain degradation index R 2 (t):

[0019]

[0020] As an optional implementation, the time domain degradation index R 1 (t) and frequency domain degradation index R 2 (t) weighted summation to obtain the comprehensive degradation trend sequence R(t): R(t) = ω 1 ·R 1 (t)+ω 2 ·R 2 (t); where ω 1 and ω 2 is the weighting coefficient; specifically:

[0021] As an optional implementation, the Autoformer model is used to predict the nonlinear component, and the ARMA model is used to predict the linear component, and each prediction result is accumulated and summed to obtain the final degradation trend prediction result.

[0022] In a second aspect, the present invention provides an integrated prediction system for degradation trend of a pumped storage unit, comprising:

[0023] A time domain health model building module is configured to obtain historical operation data and corresponding historical status data of the pumped storage unit under different working conditions, and use the data to build a time domain health model;

[0024] The frequency domain health model construction module is configured to perform modal decomposition on the historical state data, obtain the frequency domain feature vector under the health state according to the approximate entropy of each modal component, cluster the frequency domain feature vector to obtain the health center vector, and thereby construct the frequency domain health model;

[0025] The degradation index determination module is configured to obtain real-time operation data, adopt a time domain health model and a frequency domain health model, obtain a time domain degradation index and a frequency domain degradation index respectively, and obtain a comprehensive degradation trend sequence based on the two types of degradation indicators;

[0026] The integrated prediction module is configured to decompose the comprehensive degradation trend sequence, predict the obtained nonlinear component and linear component respectively, and superimpose the prediction results to obtain the final degradation trend prediction result.

[0027] In a third aspect, the present invention provides an electronic device comprising a memory and a processor, and computer instructions stored in the memory and executed on the processor, wherein when the computer instructions are executed by the processor, the method described in the first aspect is performed.

[0028] In a fourth aspect, the present invention provides a computer-readable storage medium for storing computer instructions, wherein when the computer instructions are executed by a processor, the method described in the first aspect is performed.

[0029] In a fifth aspect, the present invention provides a computer program product, comprising a computer program, which, when executed by a processor, implements the method described in the first aspect.

[0030] Compared with the prior art, the present invention has the following beneficial effects:

[0031] The present invention provides a method and system for integrated prediction of degradation trend of pumped storage units. The method considers multiple operating parameters, uses a time series convolutional network and multiple operating parameters and vibration data to construct a time domain health model, establishes a frequency domain health model based on ensemble empirical mode decomposition, approximate entropy and K-means clustering algorithm, establishes a time domain degradation index and a frequency domain degradation index according to the time domain health model and the frequency domain health model, respectively, and weights the two to calculate a comprehensive degradation trend sequence, decomposes the comprehensive degradation trend sequence using fully adaptive noise integrated empirical mode decomposition, predicts nonlinear components and linear components respectively according to different time scale characteristics of the decomposed components, and finally superimposes the predicted values ​​to obtain a final prediction result, which can more reasonably and accurately predict the degradation trend of the pumped storage unit.

[0032] In view of the shortcomings of constructing unit health models and degradation indicators in the current prediction of degradation trend of pumped-storage units, the present invention proposes to construct a comprehensive degradation sequence of pumped-storage units by combining time domain health models and frequency health models, which comprehensively considers the changing characteristics and trends of the unit's operating status in the time and frequency domains, can comprehensively reflect the degradation information of the unit, and lay a good foundation for subsequent degradation prediction.

[0033] The present invention adopts a feature extraction method, and after decomposing the original data into subsequences using fully adaptive noise integrated empirical mode decomposition, it can reduce the non-stationarity of the sequence, achieve noise reduction effect, and thus improve the prediction accuracy of the model. At the same time, according to the different time scale characteristics of the decomposed components, they are divided into nonlinear components and linear components, and the Autoformer model and ARMA model are used for prediction respectively. Compared with other models, it has higher prediction accuracy and stronger stability, making up for the deficiency that the Transformer model cannot analyze and predict long time series.

[0034] Advantages of additional aspects of the present invention will be given in part in the following description, and in part will become obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0036] Figure 1 A flow chart of an integrated prediction method for the degradation trend of a pumped storage unit provided in Example 1 of the present invention;

[0037] Figure 2 A fitting result diagram of the temporal convolutional network provided in Example 1 of the present invention;

[0038] Figure 3 A comprehensive degradation trend sequence diagram provided in Example 1 of the present invention;

[0039] Figure 4 A graph showing the fully adaptive noise integrated empirical mode decomposition results provided in Example 1 of the present invention;

[0040] Figure 5 This is a comparison chart of prediction results of different models provided in Example 1 of the present invention. DETAILED DESCRIPTION

[0041] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0042] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0043] It should be noted that the terms used herein are only for describing specific embodiments, and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that the terms "include" and "comprise" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0044] In the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other.

[0045] Example 1

[0046] like Figure 1 As shown, this embodiment provides an integrated prediction method for the degradation trend of a pumped storage unit, including:

[0047] Obtain the historical operation data and corresponding historical status data of the pumped storage unit under different working conditions, and use them to build a time domain health model;

[0048] Perform modal decomposition on historical state data, obtain the frequency domain feature vector under the healthy state according to the approximate entropy of each modal component, cluster the frequency domain feature vector to obtain the health center vector, and thus construct the frequency domain health model;

[0049] Obtain real-time operation data, use time domain health model and frequency domain health model to obtain time domain degradation index and frequency domain degradation index respectively, and obtain comprehensive degradation trend sequence based on the two types of degradation indexes;

[0050] The comprehensive degradation trend sequence is decomposed, the obtained nonlinear component and linear component are predicted respectively, and the prediction results are superimposed to obtain the final degradation trend prediction result.

[0051] In this embodiment, the collected historical operation data and corresponding historical status data of the pumped storage unit under different working conditions are obtained;

[0052] Specifically: the operating data includes water head, active power, reactive power, excitation voltage, excitation current and guide vane opening, and the status data includes vibration data.

[0053] The vibration data is used as the performance parameter for evaluating the health status of the pumped-storage unit, and the head, active power, reactive power, excitation voltage, excitation current and guide vane opening are used as operating parameters to characterize the operating conditions of the pumped-storage unit. Thus, a data set consisting of operating parameters and performance parameters is obtained.

[0054] In this embodiment, a temporal convolutional network (TCN) is used, and the historical operation data under different working conditions is used as the input of the temporal convolutional network TCN, and the corresponding historical state data, that is, the vibration data, is used as the output of the temporal convolutional network TCN, so as to effectively use the historical monitoring data to construct the time domain health model THM, which is expressed as follows:

[0055] F(t)=f(H(t),P(t),Q(t),U m (t),I m (t), G(t));

[0056] Among them, F(t) is the vibration data, H(t) is the working water head, P(t) is the active power, Q(t) is the reactive power, U m (t) is the excitation voltage, I m (t) is the excitation current, G(t) is the guide vane opening.

[0057] The fitting results of the temporal convolutional network are as follows Figure 2 shown.

[0058] In this embodiment, a frequency domain health model FHM is established based on ensemble empirical mode decomposition (EEMD), approximate entropy (ApEn) and K-means clustering algorithm;

[0059] Specifically:

[0060] The vibration data corresponding to the time domain health model THM is selected, and the EEMD algorithm is used to decompose the vibration data to obtain 6 groups of modal components;

[0061] Calculate the approximate entropy of each modal component respectively to obtain the frequency domain eigenvector L in the healthy state: L = [ApEn1, ApEn2, ..., ApEn6];

[0062] The K-means method is used to automatically cluster the frequency domain feature vectors of the unit health samples to obtain the cluster center Ω under the healthy state. The cluster center Ω is the health center vector of the frequency domain health model FHM. The health center vector value represents the frequency domain feature vector of the unit when it is operating normally. Therefore, it can be used as a health vector to evaluate the unit degradation from the frequency domain perspective.

[0063] In this embodiment, the EEMD decomposition includes the following steps:

[0064] (1) The white noise n with a set noise level is i (t) is added to the original signal x i (t) to form a new signal: x i (t) = x(t) + n i (t); where n i (t) represents the ith additive white noise sequence, x i (t) represents the additional noise signal of the i-th experiment, i=1, 2,..., M, M is the overall average number, that is, the number of times white noise is added, and its value range is 100 to 300.

[0065] (2) In order to obtain a series of IMF components c by EMD decomposition of the synthesized new signal ij (t) and the residual term r ij(t): The number of IMF components is m, among which c ij (t) is the jth component decomposed after adding white noise for the i-th time, and J is the number of IMFs.

[0066] (3) Repeat steps (2) and (3) M times and average the overall results. The result is the IMF component of the original signal x(t) obtained by EEMD decomposition:

[0067]

[0068] Among them, c j (t) is the j-th IMF of EEMD decomposition, i = 1, 2, ···, M; j = 1, 2, ···, J.

[0069] In this embodiment, the constructed time domain health model and frequency domain health model are used to construct a time domain degradation index TDI and a frequency domain degradation index FDI respectively, and the weighted sum of the two is used to calculate the comprehensive degradation trend CDI.

[0070] Specifically:

[0071] (1) Select the unit operating data in chronological order and input the operating data into the time domain health model FHM to obtain the health value Y(t) under the current operating condition. The relative error between the actual value T(t) and the health value Y(t) is taken as the time domain degradation index, and R is used as the time domain degradation index. 1 (t) represents the time domain degradation indicator sequence, specifically:

[0072]

[0073] (2) Select the corresponding vibration data, extract the frequency domain feature vector of the unit based on EEMD-ApEn, obtain the frequency domain feature vector set L(t), calculate the Euclidean distance between L(t) and the health center vector Ω, and obtain the frequency domain degradation index FDI. 2 (t) represents the frequency domain degradation indicator sequence, specifically:

[0074]

[0075] (3) TDI R 1 (t) and FDI R 2 (t) weighted summation, we get the comprehensive degradation trend sequence R(t): R(t) = ω 1 ·R 1 (t)+ω 2 ·R 2 (t); where ω 1 and ω 2 is the weighting coefficient.

[0076] In order to enhance the sensitivity of the degradation index to abnormal data, the weights are:

[0077]

[0078] The comprehensive degradation trend series is as follows Figure 3 shown.

[0079] In this embodiment, in order to improve the prediction accuracy of the degradation trend of the pumped storage unit, the comprehensive degradation trend sequence is decomposed into multiple groups of subsequences using Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN), such as Figure 4 shown.

[0080] The decomposition process specifically includes the following steps:

[0081] (1) Add adaptively decomposable Gaussian white noise W(n) to the original signal y(t) to obtain a new signal y i (t).

[0082] (2) The new signal y i (t) Perform EMD decomposition and calculate the average value to obtain the first modal component IMF 1 (n), and then calculate the residual R of the first modal component 1 (n).

[0083] (3) Let the kth modal component after EMD processing be E k (·), Gaussian white noise is added to the residual signal R 1 (n), perform EMD decomposition again to calculate E 1 (R 1 (n)+ε 1 E 1 (W i (n))) average value to obtain the second mode component.

[0084] (4) Calculate the residual component after removing the kth modal component, and continuously decompose the data to obtain the value of the k+1th modal component.

[0085] (5) Repeat the above steps until the convergence condition is reached. At this point, the original signal is decomposed into k modal components and 1 residual component.

[0086] In this embodiment, the above decomposition process obtains 6 nonlinear components and 3 linear components. According to the different time scale characteristics of the decomposed components, the Autoformer model is used to predict the nonlinear components. This is because the nonlinear part has more complex characteristic information. At the same time, the ARMA model is used to predict the linear components. Finally, the prediction results of each subsequence are accumulated and summed to obtain the final degradation trend prediction result.

[0087] Transformer cannot analyze and predict long-term time series. The Autoformer model uses the Auto-Correlation mechanism to replace the self-attention mechanism based on the Transformer model, and the model introduces a deep decomposition structure to allow the model to decompose trend components during the decoding process, which provides a basis for the sequence decomposition method of subsequent models. In the entire Autoformer model, the Auto-Correlation module and the sequence decomposition module play a major role. The Auto-Correlation mechanism and serial connection are used in the Auto-Correlation module to expand the use of information. Auto-Correlation discovers period-based correlation by calculating the autocorrelation of the sequence, and aggregates similar subsequences through time delay aggregation, making it more suitable for the prediction of long-term time series.

[0088] The ARMA model is a commonly used model in time series analysis, combining autoregression (AR) and moving average (MA). The AR part predicts the current value through the linear relationship between the current value and the values ​​of several previous time points, which can effectively capture the linear trend and periodic characteristics in the time series. The MA part reduces the impact of noise on the prediction by modeling the current and past random error terms, which helps to smooth short-term fluctuations in the time series and enables the model to better capture the underlying linear laws.

[0089] In this embodiment, the first 90% of the samples of each component are divided into a training set for model training, and the last 10% of the samples are divided into a test set for verifying the effectiveness of the model. Four evaluation indicators, namely root mean square error (RMSE), mean absolute error (MAE), mean absolute percentage error (MAPE) and determination coefficient (R2), are used to perform error analysis on the model prediction results.

[0090] The four evaluation indicators include:

[0091]

[0092]

[0093] Where: n represents the number of prediction samples; y i represents the actual value of the i-th sample; Represents the predicted value of the i-th sample; Represents the average of the actual values ​​of n predicted samples.

[0094] In order to fully demonstrate the effectiveness of the pumped storage unit degradation trend prediction method based on the comprehensive degradation index and dual data-driven model, the CEEMDAN-Autoformer-ARMA model proposed in this embodiment is compared with five prediction models, namely LSTM, Autoformer, CEEMDAN-ARMA, CEEMDAN-LSTM and CEEMDAN-Autoformer. The model prediction results are shown in Figure 2. Figure 5 As shown, Figure 5 (a)-(f) are the prediction result diagrams of LSTM, Autoformer, CEEMDAN-ARMA, CEEMDAN-LSTM and the model proposed in this embodiment, respectively, and the evaluation index results are shown in Table 1. It can be seen that the predicted value of the degradation trend of the pumped storage unit of the CEEMDAN-Autoformer-ARMA model proposed in this embodiment is basically consistent with the true value. At the same time, the predicted trend of the method proposed in this embodiment is basically the same as the trend of the true value, without leading or lagging, and has a high prediction accuracy, which proves the superiority and effectiveness of the present invention.

[0095] Table 1 Performance evaluation indicators of six models

[0096]

[0097] It can be seen from Table 1 that compared with the five comparison models, the model proposed in this embodiment has four index improvement rates P RMSE , P MAE , P MAPE and There is a significant improvement in P RMSE , P MAE , P MAPE and The minimum improvement rates are 38.60%, 31.32%, 21.23% and 4.440% respectively. The results show that the comparison model cannot accurately predict the true value, which further proves the effectiveness of the method proposed in this embodiment, and also shows that the prediction stability of the method proposed in this embodiment is better than that of the other five comparison methods.

[0098] Example 2

[0099] This embodiment provides a pumped storage unit degradation trend integrated prediction system, including:

[0100] A time domain health model building module is configured to obtain historical operation data and corresponding historical status data of the pumped storage unit under different working conditions, and use the data to build a time domain health model;

[0101] The frequency domain health model construction module is configured to perform modal decomposition on the historical state data, obtain the frequency domain feature vector under the health state according to the approximate entropy of each modal component, cluster the frequency domain feature vector to obtain the health center vector, and thereby construct the frequency domain health model;

[0102] The degradation index determination module is configured to obtain real-time operation data, adopt a time domain health model and a frequency domain health model, obtain a time domain degradation index and a frequency domain degradation index respectively, and obtain a comprehensive degradation trend sequence based on the two types of degradation indicators;

[0103] The integrated prediction module is configured to decompose the comprehensive degradation trend sequence, predict the obtained nonlinear component and linear component respectively, and superimpose the prediction results to obtain the final degradation trend prediction result.

[0104] As an optional implementation, the operating data includes water head, active power, reactive power, excitation voltage, excitation current and guide vane opening; the status data includes vibration data; a time domain health model is constructed by using a time series convolutional network with historical operating data as input and historical status data as output.

[0105] As an optional implementation, the process of obtaining the time domain degradation index using the time domain health model includes:

[0106] The real-time operation data is input into the time domain health model to obtain the health value Y(t) under the current working condition, and the relative error between the actual value T(t) and the health value Y(t) is taken as the time domain degradation index R 1 (t):

[0107] As an optional implementation method, the process of obtaining the frequency domain degradation index using the frequency domain health model includes:

[0108] The frequency domain feature vector L(t) is extracted from the status data corresponding to the real-time operation data, and the Euclidean distance between L(t) and the health center vector Ω is calculated to obtain the frequency domain degradation index R 2 (t):

[0109]

[0110] As an optional implementation, the time domain degradation index R 1 (t) and frequency domain degradation index R 2 (t) weighted summation to obtain the comprehensive degradation trend sequence R(t): R(t) = ω 1 ·R1 (t)+ω 2 ·R 2 (t); where ω 1 and ω 2 is the weighting coefficient; specifically:

[0111] As an optional implementation, the Autoformer model is used to predict the nonlinear component, and the ARMA model is used to predict the linear component, and each prediction result is accumulated and summed to obtain the final degradation trend prediction result.

[0112] It should be noted that the above modules correspond to the steps described in Example 1, and the examples and application scenarios implemented by the above modules and the corresponding steps are the same, but are not limited to the contents disclosed in the above Example 1. It should be noted that the above modules, as part of the system, can be executed in a computer system such as a set of computer executable instructions.

[0113] In further embodiments, there is also provided:

[0114] An electronic device includes a memory and a processor, and computer instructions stored in the memory and executed on the processor, wherein when the computer instructions are executed by the processor, the method described in Embodiment 1 is performed. For the sake of brevity, it will not be described in detail here.

[0115] It should be understood that in this embodiment, the processor may be a central processing unit CPU, and the processor may also be other general-purpose processors, digital signal processors DSP, application-specific integrated circuits ASIC, off-the-shelf programmable gate arrays FPGA or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0116] The memory may include a read-only memory and a random access memory, and provide instructions and data to the processor. A portion of the memory may also include a non-volatile random access memory. For example, the memory may also store information about the device type.

[0117] A computer-readable storage medium is used to store computer instructions, and when the computer instructions are executed by a processor, the method described in Example 1 is completed.

[0118] The method in Example 1 can be directly embodied as a hardware processor, or a combination of hardware and software modules in the processor. The software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory, and the processor reads the information in the memory and completes the steps of the above method in combination with its hardware. To avoid repetition, it is not described in detail here.

[0119] A computer program product includes a computer program, and when the computer program is executed by a processor, the method described in embodiment 1 is implemented.

[0120] The present invention also provides at least one computer program product tangibly stored on a non-transitory computer-readable storage medium. The computer program product includes computer executable instructions, such as instructions included in a program module, which are executed in a device on a real or virtual processor of the target to perform the process / method as described above. Typically, program modules include routines, programs, libraries, objects, classes, components, data structures, etc. that perform specific tasks or implement specific abstract data types. In various embodiments, the functions of program modules can be combined or divided between program modules as needed. Machine executable instructions for program modules can be executed in local or distributed devices. In distributed devices, program modules can be located in local and remote storage media.

[0121] The computer program code for implementing the method of the present invention can be written in one or more programming languages. These computer program codes can be provided to the processor of a general-purpose computer, a special-purpose computer or other programmable data processing device, so that the program code, when executed by the computer or other programmable data processing device, causes the function / operation specified in the flow chart and / or block diagram to be implemented. The program code can be executed completely on a computer, partially on a computer, as an independent software package, partially on a computer and partially on a remote computer or completely on a remote computer or server.

[0122] In the context of the present invention, computer program codes or related data may be carried by any appropriate carrier to enable a device, apparatus or processor to perform the various processes and operations described above. Examples of carriers include signals, computer readable media, etc. Examples of signals may include electrical, optical, radio, acoustic or other forms of propagation signals, such as carrier waves, infrared signals, etc.

[0123] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with this embodiment can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0124] Although the above describes the specific implementation mode of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without creative work are still within the scope of protection of the present invention.

Claims

1. A method for integrated prediction of degradation trend of pumped storage units, characterized in that: include: Obtain the historical operation data and corresponding historical status data of the pumped storage unit under different working conditions, and use them to build a time domain health model; Perform modal decomposition on historical state data, obtain the frequency domain feature vector under the healthy state according to the approximate entropy of each modal component, cluster the frequency domain feature vector to obtain the health center vector, and thus construct the frequency domain health model; Obtain real-time operation data, use time domain health model and frequency domain health model to obtain time domain degradation index and frequency domain degradation index respectively, and obtain comprehensive degradation trend sequence based on the two types of degradation indexes; The comprehensive degradation trend sequence is decomposed, the obtained nonlinear component and linear component are predicted respectively, and the prediction results are superimposed to obtain the final degradation trend prediction result.

2. The integrated prediction method for degradation trend of a pumped storage unit according to claim 1, characterized in that: The operating data include water head, active power, reactive power, excitation voltage, excitation current and guide vane opening; the state data include vibration data; a time domain health model is constructed by using a time series convolutional network with historical operating data as input and historical state data as output.

3. The integrated prediction method for degradation trend of a pumped storage unit according to claim 1, characterized in that: The process of obtaining the time domain degradation index using the time domain health model includes: The real-time operation data is input into the time-domain health model to obtain the health value Y(t) under the current working condition, and the relative error between the actual value T(t) and the health value Y(t) is taken as the time-domain degradation index R1(t):

4. The integrated prediction method for degradation trend of a pumped storage unit according to claim 3, characterized in that: The process of obtaining the frequency domain degradation index using the frequency domain health model includes: The frequency domain feature vector L(t) is extracted from the status data corresponding to the real-time operation data, and the Euclidean distance between L(t) and the health center vector Ω is calculated to obtain the frequency domain degradation index R2(t):

5. The integrated prediction method for degradation trend of a pumped storage unit according to claim 1, characterized in that: The time domain degradation index R1(t) and the frequency domain degradation index R2(t) are weighted and summed to obtain the comprehensive degradation trend sequence R(t): R(t) = ω1·R1(t)+ω2·R2(t); where ω1 and ω2 are weighting coefficients; specifically:

6. The integrated prediction method for degradation trend of a pumped storage unit according to claim 1, characterized in that: The Autoformer model is used to predict the nonlinear component, and the ARMA model is used to predict the linear component. The prediction results are accumulated and summed to obtain the final degradation trend prediction result.

7. An integrated prediction system for degradation trend of pumped storage units, characterized in that: include: A time domain health model building module is configured to obtain historical operation data and corresponding historical status data of the pumped storage unit under different working conditions, and use the data to build a time domain health model; The frequency domain health model construction module is configured to perform modal decomposition on the historical state data, obtain the frequency domain feature vector under the health state according to the approximate entropy of each modal component, cluster the frequency domain feature vector to obtain the health center vector, and thereby construct the frequency domain health model; The degradation index determination module is configured to obtain real-time operation data, adopt a time domain health model and a frequency domain health model, obtain a time domain degradation index and a frequency domain degradation index respectively, and obtain a comprehensive degradation trend sequence based on the two types of degradation indicators; The integrated prediction module is configured to decompose the comprehensive degradation trend sequence, predict the obtained nonlinear component and linear component respectively, and superimpose the prediction results to obtain the final degradation trend prediction result.

8. An electronic device, characterized in that: The method comprises a memory and a processor and computer instructions stored in the memory and executed on the processor, wherein when the computer instructions are executed by the processor, the method according to any one of claims 1 to 6 is completed.

9. A computer-readable storage medium, characterized in that: Used to store computer instructions, which, when executed by a processor, complete the method described in any one of claims 1 to 6.

10. A computer program product, characterized in that The invention comprises a computer program, which, when executed by a processor, implements the method according to any one of claims 1 to 6.

Citation Information

Patent Citations

  • Hybrid dynamic fault prediction method based on ARMA and ANN

    CN107085750A

  • Pumped storage unit vibration trend prediction method

    CN108875841A

  • Hydroelectric generating set degradation degree prediction method based on EEMD and LSTM

    CN111461450A

  • Pumped storage unit health state model construction and performance trend prediction method and device

    CN114880917A

  • Fault diagnosis method for rolling bearing

    CN115221930A