A method and system for predicting degradation of hydropower units

Abstract

The embodiment of the present invention provides a method and system for predicting degradation of a hydropower unit, the method comprising: training healthy state models of different swings based on the collected operating parameters and swing data of the hydropower unit under steady-state conditions; calculating the swing degradation sequences of different unit components, and fusing the swing degradation sequences of the different components to obtain a fused degradation sequence; predicting the fused swing degradation sequence to obtain a degradation trend. The embodiment of the present invention provides a method and system for predicting degradation of a hydropower unit, which respectively establishes healthy state models for the swings of different components, and calculates the fused degradation degrees of different components, which can reflect the states of different components and is more conducive to comprehensively reflecting the overall degradation degree of the unit, and utilizes the advantages of gated recurrent unit neural networks in extracting features from time series data and the stronger learning ability of deep neural networks compared to shallow networks, so that the prediction accuracy of the model is higher.

Description

Technical Field

[0001] The present invention relates to the technical field of safe operation and maintenance of hydropower units, and in particular to a method and system for predicting degradation of hydropower units. Background Art

[0002] Hydropower units play an important role in power generation, frequency regulation, peak shaving and valley filling. As the operating time increases, the units will gradually deteriorate or even fail. Obtaining the degradation sequence of hydropower units through the health state model and predicting the degradation sequence of hydropower units to obtain the degradation trend is conducive to improving the understanding of the operation and degradation laws of unit equipment, improving the management and operation and maintenance level of power station equipment, thereby generating huge economic, social and environmental benefits, and has important theoretical significance and engineering application value.

[0003] There are few studies on the prediction of degradation trend of hydropower units, most of which are based on the state change of a single vibration or swing to reflect the degradation trend of the unit. Since the hydropower unit is a complex system affected by hydraulic factors, mechanical factors, electromagnetic factors, etc., the state change of a single component is difficult to fully reflect the overall state of the unit. In addition, the artificial neural network or gray prediction methods used in existing studies have simple model structures and low prediction accuracy. Summary of the invention

[0004] The embodiment of the present invention provides a method and system for predicting degradation of a hydropower unit, which is used to solve the defects of the existing methods such as artificial neural network or grey prediction used in predicting degradation trend of hydropower units, which have simple model structure and low prediction accuracy.

[0005] In a first aspect, the present invention provides a method for predicting degradation of a hydropower unit, comprising:

[0006] Based on the collected operating parameters and swing data of the hydropower unit under steady-state conditions, the health status models with different swings are trained;

[0007] Calculating the swing degradation sequences of different unit components, and fusing the swing degradation sequences of the different components to obtain a fused degradation sequence;

[0008] The degradation trend of the fused swing degradation sequence is predicted.

[0009] Preferably, the health state models with different swings are trained based on the collected operating parameters and swing data of the hydropower unit under steady-state conditions, including:

[0010] Based on the maximum mutual information coefficient method, the correlation analysis of the collected operating parameters and swing data of the hydropower unit under steady-state conditions is carried out, and the operating parameters and swing data are normalized;

[0011] Based on a fully connected neural network, a fully connected network model is established, and the fully connected network model is trained to obtain a mapping relationship between the swing and the data parameters.

[0012] Preferably, the calculation of the swing degradation sequence of different unit components includes:

[0013] The normalized working condition parameter data and swing data after several years are used as input arrays and input into corresponding mapping relationships to obtain the health status standard values ​​of the swings of different components;

[0014] The output array is compared with the health state standard value according to the formula defined by the degradation time series to obtain the degradation time series of the hydropower unit with different swings.

[0015] Preferably, the step of fusing the swing degradation sequences of different components to obtain a fused degradation sequence includes:

[0016] The swing degradation sequences of different components are weightedly fused to obtain a fused degradation sequence.

[0017] Preferably, the step of predicting the fused swing degradation sequence to obtain a degradation trend includes:

[0018] The gated recurrent unit neural network is used to predict the fused swing degradation sequence and obtain the swing degradation trend of the hydropower unit under steady state.

[0019] Preferably, the gated recurrent unit neural network is used to predict the fused swing degradation sequence to obtain the swing degradation trend of the hydropower unit in a steady state, including:

[0020] Dividing the degradation time series into a training data set and a test data set according to a preset ratio;

[0021] Training the gated recurrent unit neural network based on the training data set;

[0022] The trained gated recurrent unit neural network is verified based on the test data set to obtain a predicted degradation trend sequence.

[0023] In a second aspect, an embodiment of the present invention provides a hydropower unit degradation prediction system, comprising:

[0024] The model training module is used to train the health status models of different swings according to the collected operating parameters and swing data of the hydropower unit under steady-state conditions;

[0025] A degradation sequence fusion module is used to calculate the swing degradation sequences of different unit components and fuse the swing degradation sequences of the different components to obtain a fused degradation sequence;

[0026] The degradation prediction module is used to predict the fusion swing degradation sequence to obtain the degradation trend.

[0027] In a third aspect, an embodiment of the present invention provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of a method for predicting degradation of a hydropower unit as provided in the first aspect above are implemented.

[0028] In a fourth aspect, an embodiment of the present invention provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of a method for predicting degradation of a hydropower unit as provided in the first aspect above.

[0029] The hydropower unit degradation prediction method and system provided in the embodiment of the present invention establish health status models for the swings of different components respectively, and calculate the fused degradation degrees of different components, which can not only reflect the states of different components, but also be more conducive to comprehensively reflecting the overall degradation degree of the unit. The advantages of the gated recurrent unit neural network in extracting features from time series data and the stronger learning ability of deep neural networks compared to shallow networks are utilized to make the prediction accuracy of the model higher. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0031] Figure 1 It is a schematic flow chart of a method for predicting degradation of a hydropower unit provided by an embodiment of the present invention;

[0032] Figure 2 A flow chart of a method for predicting degradation of a hydropower unit based on a fusion swing model under steady-state conditions provided by the present invention;

[0033] Figure 3 A structural block diagram of a gated recurrent unit neural network provided by an embodiment of the present invention;

[0034] Figure 4 A diagram showing the prediction results of a swing fusion degradation sequence provided by an embodiment of the present invention;

[0035] Figure 5 It is a schematic diagram of the structure of a hydropower unit degradation prediction system provided by an embodiment of the present invention;

[0036] Figure 6It is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0037] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0038] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0039] At present, there are few studies on the prediction of degradation trend of existing hydropower units. Most of them are based on the state change of a single vibration or swing to reflect the degradation trend of the unit. Since the hydropower unit is a complex system affected by hydraulic factors, mechanical factors, electromagnetic factors, etc., the state change of a single component is difficult to fully reflect the overall state of the unit. In addition, the artificial neural network or gray prediction methods used in existing studies have simple model structures and low prediction accuracy.

[0040] Therefore, the embodiment of the present invention provides a method and system for predicting degradation of a hydropower unit, which establishes a health state model for the swing of different components, and calculates the fusion degradation degree of different components, which can not only reflect the status of different components, but also is more conducive to comprehensively reflecting the overall degradation degree of the unit, and utilizes the advantages of the gated recurrent unit neural network in extracting time series data features and the stronger learning ability of the deep neural network than the shallow network, so that the prediction accuracy of the model is higher. The following will be explained and introduced through multiple embodiments in conjunction with the accompanying drawings.

[0041] Figure 1 FIG. 1 is a flow chart of a method for predicting degradation of a hydropower unit provided by an embodiment of the present invention. Figure 1 As shown, including:

[0042] 101. Based on the collected operating parameters and swing data of the hydropower unit under steady-state conditions, train the health status models with different swings;

[0043] 102. Calculate the swing degradation sequences of different unit components, and fuse the swing degradation sequences of the different components to obtain a fused degradation sequence;

[0044] 103. The degradation trend of the fused swing degradation sequence is predicted.

[0045] Figure 2 The present invention provides a flow chart of a method for predicting degradation of a hydropower unit based on a fusion swing model under steady-state conditions, such as Figure 2 As shown, it mainly includes the following processes:

[0046] First, in step 101, the embodiment of the present invention collects data of the hydropower unit under steady-state conditions, and then pre-processes the data to obtain health status models with different swings, such as Figure 2 As shown, it includes health status models such as upper guide X direction, upper guide Y direction, water guide X direction, and water guide Y direction.

[0047] Then in step 102, the swing degradation sequence of different unit components is calculated, such as Figure 2 As shown, the degradation time series of the upper guide X direction, the upper guide Y direction, the water guide X direction, the water guide Y direction, etc. are then fused to obtain the fused degradation sequence.

[0048] Finally, in step 103, the fused swing degradation sequence is predicted to obtain the degradation trend.

[0049] In one embodiment, the health state models with different swings are trained based on the collected operating parameters and swing data of the hydropower unit under steady-state conditions, including:

[0050] Based on the maximum mutual information coefficient method, the correlation analysis is performed on the collected operating parameters and swing data of the hydropower unit under steady-state conditions, and the operating parameters and the swing data are normalized; based on the fully connected neural network, a fully connected network model is established, and the fully connected network model is trained to obtain the mapping relationship between the swing and the data parameters.

[0051] Specifically, the hydropower unit is in a transient process during the start-up and shutdown process, and vibration parameters such as swing change dramatically, which is not conducive to the analysis of the unit status. Therefore, the steady-state operating parameters and swing data of the hydropower unit are collected 1 hour after the start-up. The operating parameters of the hydropower unit such as active power P, reactive power Q, working water head H, excitation current I, excitation voltage U, upper guide bearing oil outlet temperature T1, lower guide bearing oil outlet temperature T2, water guide bearing temperature T3 are collected every m seconds; upper guide X-direction swing F1, upper guide Y-direction swing F2, lower guide X-direction swing F3, lower guide Y-direction swing F4, water guide X-direction swing F5, water guide Y-direction swing F6 and other swing data, the size of m is selected according to actual needs;

[0052] Then, the maximum information coefficient method is used to calculate the correlation between the above working condition parameters and each swing, and the working condition parameters with the largest correlation coefficient with the swing are selected as the input of the health state model;

[0053] Take N groups of initial fault-free data of the unit after correlation analysis and screening, and normalize the operating parameter data and the swing data to the interval [0, 1]. Specifically, in the embodiment of the present invention, take 100,000 groups of initial fault-free data of the unit after correlation analysis and screening, and normalize the operating parameter data and the swing data to the interval [0, 1]. The normalization calculation formula is:

[0054]

[0055] where x i is the collected data, x min is the minimum value of the collected data, x max is the maximum value in the collected data.

[0056] The normalized data is matrixed: for each swing, the normalized operating parameter data is taken as input, and the normalized swing data is taken as output, to obtain a plurality of input arrays P1 with a columns and N rows and an output array T1 with 1 column and N rows, where a is the number of operating parameters after correlation analysis and screening, and in the embodiment of the present invention, six input arrays P1 with 6 columns and 100,000 rows and an output array T1 with 1 column and 100,000 rows are obtained;

[0057] Then build a multi-layer fully connected network, which includes an input layer, multiple hidden layers and an output layer. The output of the previous hidden layer is used as the input of the next hidden layer, and the output of the last hidden layer is used as the input of the output layer. The input of the fully connected network is the operating parameters of the unit, and the output is the swing data. The hidden layer uses the Sigmoid activation function, and the output layer uses the linear activation function. Specifically, in an embodiment of the present invention, the fully connected network includes an input layer, 7 hidden layers and an output layer. The output of the previous hidden layer is used as the input of the next hidden layer, and the output of the last hidden layer is used as the input of the output layer. The number of neurons in the hidden layer is 128, 64, 32, 16, 8, 4, and 2, respectively. The input of the fully connected network is the operating parameters of the unit, and the output is each swing. The hidden layer uses the Sigmoid activation function, and the output layer uses the linear activation function; wherein the Sigmoid function formula is:

[0058] The weights of the fully connected network are randomly initialized, and the size of a training sample and the number of training rounds are specified. Specifically, in the embodiment of the present invention, the size of a training sample is 128, and the number of training rounds is 100;

[0059] Take the first R1% rows of the input array P1 as the training input data set, take the first R1% rows of the output array T1 as the training reference output data set, and use the Adam gradient descent algorithm to train the multi-layer fully connected network, where R1 is the preset value. After the fully connected neural network is trained, the working condition parameters under the healthy state are obtained: active power P n , reactive power Q n , working head H n , Excitation current I n , Excitation voltage U n , related temperature T n With swing F n The mapping relationship between them is:

[0060] F n (t) = f(P n (t),Q n (t),H n (t),I n (t),U n (t),T n (t))

[0061] Specifically in the embodiment of the present invention, R1 is 80%.

[0062] Take the last (100-R1)% rows of the input array P1 as the test input data set, input it into the fully connected network, get the corresponding swing health state standard value f(t), and compare it with the last (100-R1)% rows of the output array T1, and calculate the average relative error MAPE, and find that the swing health state standard value is close to the measured value. The MAPE calculation formula is as follows:

[0063]

[0064] Wherein, N is the data length, r(t) is the measured value of the swing of the hydropower unit at the operating time t, and f(t) is the standard value output by the swing health status model.

[0065] Specifically in the embodiment of the present invention, R1 is 80%.

[0066] In one embodiment, the step of calculating the swing degradation sequence of different machine group components includes:

[0067] The normalized working condition parameter data and swing data after several years are used as input arrays and input into corresponding mapping relationships to obtain the health status standard values ​​of the swings of different components;

[0068] The output array is compared with the health state standard value according to the formula defined by the degradation time series to obtain the degradation time series of the hydropower unit with different swings.

[0069] Specifically, M groups of operating parameters and swing data after y years are taken, and normalized to the interval [0,1]. Specifically, in the embodiment of the present invention, 200,000 groups of operating parameters and swing data from November 2018 to May 2019 are taken;

[0070] The normalized working condition parameter data is used as input, and different swing data is taken as output, so as to obtain a plurality of input arrays P2 with a columns and M rows and an output array T2 with 1 column and M rows. Specifically, in the embodiment of the present invention, six input arrays P2 with 6 columns and 200,000 rows and an output array T2 with 1 column and 200,000 rows are obtained;

[0071] Substitute each input array P2 into the established multi-layer fully connected network, according to the mapping relationship F n (t) = f(P n (t),Q n (t),H n (t),I n (t),U n (t),T n (t)), and obtain the standard value of the swing under the working condition parameters after y years F n (t), and compared with the measured value of the unit swing R after y years n (t) to obtain the degradation time series with different swings. The D n (t) is defined as follows:

[0072]

[0073] Among them, K n Indicates the corresponding alarm thresholds for different swings, and t is the unit operation time.

[0074] In one embodiment, fusing the swing degradation sequences of different components to obtain a fused degradation sequence includes:

[0075] The swing degradation sequences of different components are weightedly fused to obtain a fused degradation sequence.

[0076] Specifically, the swing degradation sequences D1(t), D2(t)…D n (t) are weightedly fused to obtain the fused swing degradation sequence D(t).

[0077] In one embodiment, the step of predicting the fused swing degradation sequence to obtain a degradation trend includes:

[0078] The gated recurrent unit neural network is used to predict the fused swing degradation sequence and obtain the swing degradation trend of the hydropower unit under steady state.

[0079] In one embodiment, the gated recurrent unit neural network is used to predict the fused swing degradation sequence to obtain the swing degradation trend of the hydropower unit in a steady state, including:

[0080] Dividing the degradation time series into a training data set and a test data set according to a preset ratio;

[0081] Training the gated recurrent unit neural network based on the training data set;

[0082] The trained gated recurrent unit neural network is verified based on the test data set to obtain a predicted degradation trend sequence.

[0083] Specifically, the fused degradation sequence is processed as the input and output data of the prediction model, and L data from the i-th degradation degree to the i+L-1-th degradation degree are taken to form an input data, and the L+1-th degradation degree is taken to form an output data, and an input array P3 of len-L rows and L columns and an output array T3 of len-L rows and 1 column are obtained, where L is a preset value, i ranges from 1 to len-L, and len-L is the length of the time series. Specifically, in the embodiment of the present invention, L is 15 and len is 200000;

[0084] Normalize the input and output data, and normalize the input array P3 and the output array T3 to the interval [0,1] respectively;

[0085] Build a deep neural network based on Gate Recurrent Unit (GRU), Figure 3 A structural block diagram of a gated recurrent unit neural network provided by an embodiment of the present invention, such as Figure 3As shown, the gated recurrent unit neural network includes multiple gated recurrent unit network layers and multiple fully connected network layers, the output of the previous gated recurrent unit network layer is used as the input of the next gated recurrent unit, and the output of the last gated recurrent unit is used as the input of the fully connected network layer, and finally the prediction result is output through an output layer with a neuron, the fully connected layer uses the ReLU activation function, and the output layer uses the linear activation function. Specifically in an embodiment of the present invention, the gated recurrent unit neural network includes 2 gated recurrent unit network layers and 3 fully connected network layers, the output of the previous gated recurrent unit network layer is used as the input of the next gated recurrent unit, and the output of the last gated recurrent unit is used as the input of the fully connected network layer, and finally the prediction result is output through an output layer with a neuron, the number of neurons in the two gated recurrent unit network layers is 64 and 32, the number of neurons in the fully connected network layer is 16, 8, and 4, the fully connected layer uses the ReLU activation function, and the output layer uses the linear activation function, wherein the formula of the ReLU activation function is: g(x)=max(0,x);

[0086] The weights of the network are randomly initialized, and the size of a training sample and the number of training rounds are specified. The constructed GRU neural network is trained, and the first R3% rows of the input array P3 are taken as the input data set, and the first R3% rows of the output array T3 are taken as the output data set. The GRU deep neural network is trained using the Adam algorithm, where R3 is a preset value. Specifically, in an embodiment of the present invention, the size of a training sample is specified to be 32, and the number of training rounds is 300. The constructed GRU neural network is trained, and 80% rows of the input array P3 are taken as the input data set, and the first 80% rows of the output array T3 are taken as the output data set;

[0087] The trained GRU neural network is used to predict the test data to obtain the degradation trend. The latter (100-R3)% of the input array P3 is taken as the test input data set and input into the trained GRU neural network to obtain the predicted output of the model, and then it is denormalized to obtain the predicted degradation trend sequence. The mean absolute error MAE, mean relative error MAPE, and root mean square error RMSE between the predicted results and the actual degradation trend are calculated. The calculation formulas for each error are as follows:

[0088]

[0089]

[0090]

[0091] Among them, x i is the true value, is the prediction result, and n is the length of the sequence.

[0092] Specifically in the embodiment of the present invention, Figure 4 The swing fusion degradation sequence prediction result diagram provided by the embodiment of the present invention is as follows: Figure 4 As shown, the unit degradation degree is small and is still in a healthy state, but the degradation degree tends to increase; the mean absolute error MAE between the calculated prediction results and the actual degradation trend is 0.0041, the average relative error MAPE is 4.4%, and the root mean square error RMSE is 0.0046.

[0093] In one embodiment, Figure 5 Schematic diagram of a hydropower unit degradation prediction system provided by an embodiment of the present invention. Figure 5 As shown, it includes: a model training module 501, a degradation sequence fusion module 502 and a degradation prediction module 503, wherein:

[0094] The model training module 501 is used to train the health state models of different swings according to the collected operating parameters and swing data of the hydropower unit under steady-state conditions;

[0095] The degradation sequence fusion module 502 is used to calculate the swing degradation sequences of different unit components, and fuse the swing degradation sequences of the different components to obtain a fused degradation sequence;

[0096] The degradation prediction module 503 is used to predict the fused swing degradation sequence to obtain the degradation trend.

[0097] For details on how to use the model training module 501, the degradation sequence fusion module 502 and the degradation prediction module 503 to predict the degradation of the hydropower unit, please refer to Figure 1 The embodiments shown are for reference only and the embodiments of the present invention will not be described in detail herein.

[0098] In one embodiment, based on the same concept, an embodiment of the present invention further provides an electronic device, Figure 6 An example of a structural diagram of an electronic device is shown in FIG. Figure 6 As shown, the server may include: a processor 601, a communications interface 602, a memory 603 and a bus 604, wherein the processor 601, the communications interface 602 and the memory 603 communicate with each other through the bus 604. The processor 601 may call the logic instructions in the memory 603 to execute the following hydropower unit degradation prediction method: according to the collected hydropower unit operating parameters and swing data under steady-state conditions, train the health state models with different swings; calculate the swing degradation sequences of different unit components, and fuse the swing degradation sequences of the different components to obtain a fused degradation sequence; predict the fused swing degradation sequence to obtain the degradation trend.

[0099] In one embodiment, based on the same concept, an embodiment of the present invention further provides a non-transitory computer-readable storage medium, wherein the non-transitory computer-readable storage medium stores computer instructions, wherein the computer instructions enable the computer to execute the hydropower unit degradation prediction method provided by the above-mentioned method embodiments, for example, including: training health status models with different swings based on the collected operating parameters and swing data of the hydropower unit under steady-state conditions; calculating the swing degradation sequences of different unit components, and fusing the swing degradation sequences of the different components to obtain a fused degradation sequence; predicting the fused swing degradation sequence to obtain a degradation trend.

[0100] Based on the same technical concept, an embodiment of the present application also provides a computer program, which, when executed by a main control device, is used to implement the above method embodiment.

[0101] The program may be stored in whole or in part on a storage medium packaged together with the processor, or may be stored in whole or in part on a memory not packaged together with the processor.

[0102] Based on the same technical concept, the embodiment of the present application further provides a processor, which is used to implement the above method embodiment. The above processor can be a chip.

[0103] The various embodiments of the present invention can be combined arbitrarily to achieve different technical effects.

[0104] Those skilled in the art can understand that to implement all or part of the processes in the above-mentioned embodiments, the processes can be completed by computer programs to instruct related hardware, and the programs can be stored in computer-readable storage media. When the programs are executed, they can include the processes of the above-mentioned method embodiments. The aforementioned storage media include: ROM or random access memory RAM, magnetic disk or optical disk and other media that can store program codes.

[0105] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for predicting degradation of a hydropower unit, characterized in that: include: Based on the collected operating parameters and swing data of the hydropower unit under steady-state conditions, the health state models with different swings are trained; the swing degradation sequences of different unit components are calculated, and the swing degradation sequences of different components are fused to obtain the fused degradation sequence; The degradation trend of the fused swing degradation sequence is predicted; The step of predicting the degradation trend of the fused swing degradation sequence comprises: using a gated recurrent unit neural network to predict the degradation trend of the fused swing degradation sequence to obtain the swing degradation trend of the hydropower unit in a steady state; The method of using a gated recurrent unit neural network to predict the fused swing degradation sequence to obtain the swing degradation trend of the hydropower unit under steady state includes: dividing the degradation time series into a training data set and a test data set according to a preset ratio; training the gated recurrent unit neural network based on the training data set; verifying the trained gated recurrent unit neural network based on the test data set to obtain a predicted degradation trend sequence; training a health state model with different swings based on the collected working parameters and swing data of the hydropower unit under steady state conditions, including: performing a correlation analysis on the collected working parameters and swing data of the hydropower unit under steady state conditions based on the maximum mutual information coefficient method, and comparing the working parameters with the swing data. The data is normalized; based on a fully connected neural network, a fully connected network model is established, and the fully connected network model is trained to obtain a mapping relationship between the swing and the data parameters; the calculation of the swing degradation sequence of different unit components includes: taking the normalized operating parameter data and the swing data after several years as input arrays, respectively inputting them into the corresponding mapping relationships to obtain the health status standard values ​​of the swings of different components; comparing the output array with the health status standard value according to the formula defined by the degradation time series to obtain the degradation time series of different swings of the hydropower unit; the fusion of the swing degradation sequences of the different components to obtain a fused degradation sequence includes: weighted fusion of the swing degradation sequences of different components to obtain a fused degradation sequence.

2. A hydropower unit degradation prediction system, characterized in that: include: The model training module is used to train the health status models of different swings according to the collected operating parameters and swing data of the hydropower unit under steady-state conditions; The degradation sequence fusion module is used to calculate the swing degradation sequences of different unit components and fuse the swing degradation sequences of different components to obtain a fused degradation sequence; The degradation prediction module is used to predict the degradation trend of the fused swing degradation sequence; The method of predicting the degradation trend of the fused swing degradation sequence comprises: using a gated recurrent unit neural network to predict the degradation trend of the fused swing degradation sequence to obtain the swing degradation trend of the hydropower unit under steady state; using a gated recurrent unit neural network to predict the degradation trend of the fused swing degradation sequence to obtain the swing degradation trend of the hydropower unit under steady state comprises: dividing the degradation time series into a training data set and a test data set according to a preset ratio; training the gated recurrent unit neural network based on the training data set; verifying the trained gated recurrent unit neural network based on the test data set to obtain a predicted degradation trend sequence; training the health status models of different swings according to the collected operating parameters and swing data of the hydropower unit under steady state conditions comprises: based on the maximum mutual information coefficient method, the collected A correlation analysis is performed on the operating parameters and swing data of the hydropower unit under steady-state conditions, and the operating parameters and the swing data are normalized; based on a fully connected neural network, a fully connected network model is established, and the fully connected network model is trained to obtain a mapping relationship between the swing and the data parameters; the calculation of the swing degradation sequence of different unit components includes: taking the normalized operating parameter data and the swing data after several years as input arrays, respectively inputting them into corresponding mapping relationships to obtain the healthy state standard values ​​of the swings of different components; comparing the output array with the healthy state standard value according to the formula defined by the degradation time series to obtain the degradation time series of different swings of the hydropower unit; the fusion of the swing degradation sequences of the different components to obtain a fused degradation sequence includes: weighted fusion of the swing degradation sequences of different components to obtain a fused degradation sequence.

3. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of the hydropower unit degradation prediction method as claimed in claim 1 are implemented.

4. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for predicting degradation of a hydropower unit as claimed in claim 1 are implemented.

Images