A method, device, equipment, storage medium and program product for evaluating and predicting the deterioration trend of a light-weight hydroelectric generating set

Abstract

The method, device, equipment, storage medium and program product provided by the embodiments of the present disclosure can realize multi-position comprehensive degradation evaluation, improve the updating efficiency of the health model and the prediction model, and guarantee the long-term accuracy of online monitoring and prediction.

Description

Technical Field

[0001] This disclosure relates to the field of industrial intelligent technology, and in particular to a method, apparatus, equipment, storage medium and program product for assessing and predicting the deterioration trend of lightweight hydropower units. Background Technology

[0002] Hydropower, as a clean, sustainable, and flexible high-quality energy source, plays an increasingly important role in the transformation of the power system. Currently, the focus of my country's hydropower industry is gradually shifting from construction to operation and maintenance. As the operating time of hydropower units increases, their operational condition gradually deteriorates, with most defects developing from latent hidden dangers. Therefore, various types of sensors are deployed at key functional links of hydropower units to monitor data reflecting their operational status, providing data support for ensuring the safe operation of hydropower units.

[0003] Existing methods often reflect the deterioration status of hydropower units based on the deterioration trend of a single monitoring point, resulting in an incomplete deterioration trend. Assessing the deterioration trends of each main functional chain component and then constructing the overall deterioration trend of the hydropower unit based on this assessment is a more feasible solution. However, hydropower units have numerous key monitoring points and sensor channels, requiring massive computational resources to construct health models. Furthermore, in the deterioration trend prediction stage, deep prediction models have too many parameters, leading to excessive time and space complexity. More importantly, in practical engineering applications, the vibration and sway monitoring data of hydropower units are affected by seasonal factors, necessitating regular updates to health and prediction models to ensure the accuracy of deterioration trend assessment and prediction. The high computational resource consumption of existing methods becomes even more pronounced in these cases.

[0004] Therefore, the existing degradation trend assessment and prediction methods have not been effective when applied in practice. Problems such as excessive system resource consumption and excessive disk usage often occur. The health model and prediction model consume too many system resources when updating, causing related functions to crash and even causing serious chain reactions. Summary of the Invention

[0005] To solve the above-mentioned technical problems, or at least partially solve them, this disclosure provides a method, apparatus, equipment, storage medium, and program product for assessing and predicting the deterioration trend of lightweight hydropower units.

[0006] This disclosure provides a method for assessing and predicting the deterioration trend of lightweight hydropower units. The method includes: The operating parameters of the hydropower unit and the status monitoring quantities of each main functional chain component of the hydropower unit are collected. The operating parameters of the hydropower unit are input into a pre-trained operating parameter extraction network. The operating parameters of the hydropower unit are dimensionality reduced to obtain operating feature information. The operating feature information is input into a preset lightweight hydropower unit health model for each main functional chain component of the hydropower unit to obtain the theoretical state prediction quantity of each main functional chain component of the hydropower unit during the evaluation period. The lightweight hydropower unit health model is based on a lightweight gradient boosting mechanism. Based on the mapping relationship, the difference between the state monitoring quantity and the theoretical state prediction quantity is compared, the deterioration degree of each main functional chain part of the hydropower unit during the evaluation period is generated, the deterioration degree of each main functional chain part of the hydropower unit is analyzed, and the deterioration trend sequence of the hydropower unit during the evaluation period is generated based on the analysis results. The deterioration trend sequence of the hydropower unit is input into a preset deterioration trend prediction model to reconstruct the deterioration trend sequence of the hydropower unit, and the future deterioration trend of the hydropower unit is predicted based on the real-time reconstruction results.

[0007] The method provided in this disclosure involves collecting operating parameters of a hydropower unit and state monitoring data of various main functional chain components of the hydropower unit. The operating parameters are input into a pre-trained operating parameter extraction network. Dimensionality reduction is performed on the operating parameters to obtain operating feature information. This feature information is then input into a pre-defined lightweight hydropower unit health model for each main functional chain component of the hydropower unit to obtain theoretical state predictions for each main functional chain component of the hydropower unit during the evaluation period. This includes: Collect the operating parameters of the hydropower unit and the status monitoring data of each main functional chain component of the hydropower unit during the period to be evaluated. The hydropower unit operating parameters for the period to be evaluated are input into a pre-trained operating parameter extraction network. The operating parameter extraction network performs dimensionality reduction on the hydropower unit operating parameters to determine the operating characteristic information for the period to be evaluated. The operating condition characteristic information is input into a preset lightweight hydropower unit health model. Based on the mapping relationship between the hydropower unit operating condition parameters and corresponding state monitoring quantities of each main functional chain part of the hydropower unit in the lightweight hydropower unit health model, the theoretical state prediction quantities of each main functional chain part of the hydropower unit during the period to be evaluated are output.

[0008] The method provided in this disclosure, wherein the preset operating condition parameter extraction network includes: Construct an initial encoder with a preset number of layers, decreasing the number of neurons in each layer from the shallowest to the deepest, set an activation function for each layer, and transpose the parameters of each layer of the initial encoder to obtain an initial decoder with the same structure as the initial encoder. The initial decoder and the initial encoder are combined to obtain the operating condition parameter extraction network.

[0009] The method provided in this disclosure compares the difference between the monitored state quantity and the theoretical state prediction quantity based on the mapping relationship, generates the deterioration degree of each main functional chain component of the hydropower unit during the evaluation period, analyzes the deterioration degree of each main functional chain component of the hydropower unit, and generates a deterioration trend sequence of the hydropower unit during the evaluation period based on the analysis results, including: Based on the mapping relationship, the difference between the state monitoring quantity and the theoretical state prediction quantity of each main functional chain part of the hydropower unit is compared, and the deterioration degree of each main functional chain part of the hydropower unit to be evaluated during the corresponding evaluation period is generated. The standard deviation is used to measure the variability of the deterioration degree of each main functional chain component of the hydropower unit during the evaluation period, and the correlation coefficient is used to measure the conflict of the deterioration degree of each main functional chain component of the hydropower unit during the evaluation period. The comprehensive weights of each main functional chain component of the hydropower unit are calculated based on the variability and conflict, and the deterioration trend sequence of the hydropower unit for the evaluation period is generated by combining the CRITIC method with the comprehensive weights.

[0010] The method provided in this disclosure involves inputting the hydropower unit deterioration trend sequence into a preset deterioration trend prediction model, reconstructing the hydropower unit deterioration trend sequence, and predicting the future deterioration trend of the hydropower unit based on the real-time reconstruction results, including: The hydropower unit deterioration trend sequence is input into a preset deterioration trend prediction model. A preset function is used to determine the prediction correlation steps of the hydropower unit deterioration trend sequence. The hydropower unit deterioration trend sequence is then reconstructed based on the prediction correlation steps to obtain real-time reconstruction results. Based on the first sample input in the real-time reconstruction result, several sequence feature learning sub-models in the degradation trend prediction model are used to perform a weighted average of several prediction results according to the sub-model weights corresponding to the several sequence feature learning sub-models, and output the prediction value of the first future time point. The rolling prediction of all samples in the real-time reconstruction results yields the predicted value of the corresponding sample, and the future deterioration trend of the hydropower unit is obtained based on all the predicted values.

[0011] The method provided in this disclosure, wherein the preset degradation trend prediction model includes: Obtain historical deterioration trend sequences of hydropower units; The relevant number of steps of the historical hydropower unit deterioration trend sequence is determined by a preset function, the historical hydropower unit deterioration trend sequence is reconstructed according to the relevant number of steps, and sample-label pairs are generated according to the reconstruction results. The sample-label pairs are divided into training set and test set based on a preset percentage. The training set is resampled several times with replacement using the Bagging method to obtain several sub-training sample sets with the same number of samples as the training set. Sequence feature learning sub-models are constructed for each sub-training sample set and the models are trained. Calculate the mean squared error of each sequence feature learning sub-model on the corresponding sub-training sample set, and determine the sub-model weight of each sequence feature learning sub-model based on the mean squared error. Then, perform weighted integration of each sequence feature learning sub-model based on the sub-model weight to obtain the degradation trend prediction model. The accuracy of the degradation trend prediction model is verified using the test set. If the preset accuracy requirement is met, the degradation trend prediction model is output.

[0012] This disclosure also provides a lightweight hydropower unit degradation trend assessment and prediction device, the device comprising: The data acquisition module is used to collect the operating parameters of the hydropower unit and the status monitoring data of each main functional chain component of the hydropower unit. The operating parameters of the hydropower unit are input into a pre-trained operating parameter extraction network, and the operating parameters of the hydropower unit are dimensionality reduced to obtain operating feature information. The operating feature information is input into a preset lightweight hydropower unit health model for each main functional chain component of the hydropower unit to obtain the theoretical state prediction of each main functional chain component of the hydropower unit during the evaluation period. The lightweight hydropower unit health model is based on a lightweight gradient boosting mechanism. The analysis module is used to compare the difference between the state monitoring quantity and the theoretical state prediction quantity based on the mapping relationship, generate the degree of deterioration of each main functional chain part of the hydropower unit during the period to be evaluated, analyze the degree of deterioration of each main functional chain part of the hydropower unit, and generate the deterioration trend sequence of the hydropower unit during the period to be evaluated based on the analysis results. The prediction module is used to input the deterioration trend sequence of the hydropower unit into a preset deterioration trend prediction model, reconstruct the deterioration trend sequence of the hydropower unit, and predict the future deterioration trend of the hydropower unit based on the real-time reconstruction results.

[0013] The device provided in this disclosure, wherein the acquisition module is specifically used for: Collect the operating parameters of the hydropower unit and the status monitoring data of each main functional chain component of the hydropower unit during the period to be evaluated. The hydropower unit operating parameters for the period to be evaluated are input into a pre-trained operating parameter extraction network. The operating parameter extraction network performs dimensionality reduction on the hydropower unit operating parameters to determine the operating characteristic information for the period to be evaluated. The operating condition characteristic information is input into a preset lightweight hydropower unit health model. Based on the mapping relationship between the hydropower unit operating condition parameters and corresponding state monitoring quantities of each main functional chain part of the hydropower unit in the lightweight hydropower unit health model, the theoretical state prediction quantities of each main functional chain part of the hydropower unit during the period to be evaluated are output.

[0014] The device provided in this disclosure, wherein the acquisition module is specifically used for: Construct an initial encoder with a preset number of layers, decreasing the number of neurons in each layer from the shallowest to the deepest, set an activation function for each layer, and transpose the parameters of each layer of the initial encoder to obtain an initial decoder with the same structure as the initial encoder. The initial decoder and the initial encoder are combined to obtain the operating condition parameter extraction network.

[0015] The analysis module in the apparatus provided in this disclosure is specifically used for: Based on the mapping relationship, the difference between the state monitoring quantity and the theoretical state prediction quantity of each main functional chain part of the hydropower unit is compared, and the deterioration degree of each main functional chain part of the hydropower unit to be evaluated during the corresponding evaluation period is generated. The standard deviation is used to measure the variability of the deterioration degree of each main functional chain component of the hydropower unit during the evaluation period, and the correlation coefficient is used to measure the conflict of the deterioration degree of each main functional chain component of the hydropower unit during the evaluation period. The comprehensive weights of each main functional chain component of the hydropower unit are calculated based on the variability and conflict, and the deterioration trend sequence of the hydropower unit for the evaluation period is generated by combining the CRITIC method with the comprehensive weights.

[0016] The apparatus provided in this disclosure, wherein the prediction module is specifically used for: The hydropower unit deterioration trend sequence is input into a preset deterioration trend prediction model. A preset function is used to determine the prediction correlation steps of the hydropower unit deterioration trend sequence. The hydropower unit deterioration trend sequence is then reconstructed based on the prediction correlation steps to obtain real-time reconstruction results. Based on the first sample input in the real-time reconstruction result, several sequence feature learning sub-models in the degradation trend prediction model are used to perform a weighted average of several prediction results according to the sub-model weights corresponding to the several sequence feature learning sub-models, and output the prediction value of the first future time point. The rolling prediction of all samples in the real-time reconstruction results yields the predicted value of the corresponding sample, and the future deterioration trend of the hydropower unit is obtained based on all the predicted values.

[0017] The apparatus provided in this disclosure, wherein the prediction module is specifically used for: Obtain historical deterioration trend sequences of hydropower units; The relevant number of steps of the historical hydropower unit deterioration trend sequence is determined by a preset function, the historical hydropower unit deterioration trend sequence is reconstructed according to the relevant number of steps, and sample-label pairs are generated according to the reconstruction results. The sample-label pairs are divided into training set and test set based on a preset percentage. The training set is resampled several times with replacement using the Bagging method to obtain several sub-training sample sets with the same number of samples as the training set. Sequence feature learning sub-models are constructed for each sub-training sample set and the models are trained. Calculate the mean squared error of each sequence feature learning sub-model on the corresponding sub-training sample set, and determine the sub-model weight of each sequence feature learning sub-model based on the mean squared error. Then, perform weighted integration of each sequence feature learning sub-model based on the sub-model weight to obtain the degradation trend prediction model. The degradation trend prediction was verified using the test set.

[0018] This disclosure also provides an electronic device, comprising: a processor; a memory for storing executable instructions of the processor; the processor being configured to read the executable instructions from the memory and execute the instructions to implement the lightweight hydropower unit degradation trend assessment and prediction method provided in this disclosure.

[0019] This disclosure also provides a computer-readable storage medium storing a computer program for executing the lightweight hydropower unit degradation trend assessment and prediction method provided in this disclosure.

[0020] This disclosure also provides a computer program product, including a computer program / instruction, which is executed by a processor using the lightweight hydropower unit degradation trend assessment and prediction method provided in this disclosure.

[0021] The technical solution provided in this disclosure has the following advantages compared with the prior art: The lightweight hydropower unit degradation trend assessment and prediction method provided in this disclosure collects hydropower unit operating parameters and condition monitoring quantities, extracts operating feature information through an operating parameter extraction network, inputs the lightweight hydropower unit health model of each main functional chain part of the hydropower unit to obtain theoretical state prediction quantities, calculates the difference between the condition monitoring quantities and the theoretical state prediction quantities, generates the degradation degree of each main functional chain part of the hydropower unit and constructs the hydropower unit degradation trend sequence, and finally uses the degradation trend prediction model to reconstruct the hydropower unit degradation trend sequence and predict future degradation trends. This achieves comprehensive degradation assessment of multiple parts, significantly reduces the model's computational resource consumption and storage requirements, improves the update efficiency of the health model and prediction model, enhances system operation stability, and ensures the long-term accuracy of online monitoring and prediction. Attached Figure Description

[0022] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. Throughout the drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are schematic, and the originals and elements are not necessarily drawn to scale.

[0023] Figure 1 A flowchart illustrating the method for assessing and predicting the deterioration trend of lightweight hydropower units provided in this embodiment of the disclosure; Figure 2 This is a schematic diagram of the deep autoencoder network structure established in the embodiments of this disclosure; Figure 3 This is a schematic diagram of the degradation trend sequence of four key parts generated in an embodiment of the present invention; Figure 4 This is a schematic diagram of the unit deterioration trend sequence generated in an embodiment of the present invention; Figure 5 This is a schematic diagram of the degradation trend prediction model in an embodiment of the present invention; Figure 6 This is a graph showing the predicted performance of the degradation trend prediction model in this embodiment of the invention. Figure 7 A schematic diagram of the structure of the lightweight hydropower unit degradation trend assessment and prediction device provided in the embodiments of this disclosure; Figure 8 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this disclosure. Detailed Implementation

[0024] Embodiments of this disclosure will now be described in more detail with reference to the accompanying drawings. While some embodiments of this disclosure are shown in the drawings, it should be understood that this disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of this disclosure. It should be understood that the accompanying drawings and embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of protection of this disclosure.

[0025] It should be understood that the steps described in the method embodiments of this disclosure may be performed in different orders and / or in parallel. Furthermore, the method embodiments may include additional steps and / or omit the steps shown. The scope of this disclosure is not limited in this respect.

[0026] The term "comprising" and its variations as used herein are open-ended inclusions, meaning "including but not limited to". The term "based on" means "at least partially based on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". Definitions of other terms will be given in the description below.

[0027] It should be noted that the concepts of "first" and "second" mentioned in this disclosure are used only to distinguish different devices, modules or units, and are not used to limit the order of functions performed by these devices, modules or units or their interdependencies.

[0028] It should be noted that the terms "a" and "a plurality of" used in this disclosure are illustrative rather than restrictive, and those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".

[0029] The names of messages or information exchanged between multiple devices in the embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of such messages or information.

[0030] To address the aforementioned issues, this disclosure provides a method for assessing and predicting the deterioration trend of lightweight hydropower units. The method will be described below with reference to specific embodiments.

[0031] Figure 1 This is a flowchart illustrating a method for assessing and predicting the deterioration trend of a lightweight hydropower unit according to an embodiment of this disclosure. The method can be executed by a device for assessing and predicting the deterioration trend of a lightweight hydropower unit, wherein the device can be implemented using software and / or hardware, and is generally integrated into an electronic device.

[0032] Example 1: This embodiment of the present disclosure provides a method for assessing and predicting the deterioration trend of lightweight hydropower units, the method comprising: S101: Collect the operating parameters of the hydropower unit and the status monitoring data of each main functional chain component of the hydropower unit. Input the operating parameters of the hydropower unit into a pre-trained operating parameter extraction network. Reduce the dimensionality of the operating parameters of the hydropower unit to obtain operating feature information. Input the operating feature information into the preset lightweight hydropower unit health model of each main functional chain component of the hydropower unit to obtain the theoretical state prediction of each main functional chain component of the hydropower unit during the evaluation period. The lightweight hydropower unit health model is based on a lightweight gradient boosting mechanism. S102: Based on the mapping relationship, compare the difference between the state monitoring quantity and the theoretical state prediction quantity, generate the deterioration degree of each main functional chain part of the hydropower unit during the period to be evaluated, analyze the deterioration degree of each main functional chain part of the hydropower unit, and generate the deterioration trend sequence of the hydropower unit during the period to be evaluated based on the analysis results. S103: Input the hydropower unit deterioration trend sequence into a preset deterioration trend prediction model, reconstruct the hydropower unit deterioration trend sequence, and predict the future deterioration trend of the hydropower unit based on the real-time reconstruction results.

[0033] In this embodiment, the hydropower unit operating parameters refer to the external and internal condition parameters that describe the overall operating environment and state of the hydropower unit. Examples include active power, head, guide vane opening, excitation current, excitation voltage, and stator coil temperature.

[0034] In this embodiment, the main functional chain components refer to the key physical components that constitute the core conversion chain of hydropower unit from water energy to mechanical energy to electrical energy. Based on equipment ledgers, inspection locations, and condition monitoring technical guidelines, several main functional chain components that affect the safe and stable operation of the unit are identified.

[0035] In this embodiment, the state monitoring quantity refers to the physical quantity collected by sensors directly installed on the main functional chain parts, which reflects the physical state of those parts. For example, vibration, sway, etc., at each main functional chain part.

[0036] In this embodiment, data cleaning is performed by removing abnormal monitoring values ​​and interpolating missing values. This step completes the data collection and cleaning for the healthy operation phase of each main functional chain component of the unit.

[0037] In this embodiment, the operating condition feature information refers to the output of the operating condition parameter extraction network, which is a low-dimensional core feature vector obtained after dimensionality reduction, noise reduction, and feature extraction of the original hydropower unit operating condition parameters.

[0038] In this embodiment, the input of the lightweight hydropower unit health model is the operating condition feature information (i.e., the hydropower unit operating condition parameters after dimensionality reduction), and the output is the theoretical state prediction.

[0039] In this embodiment, the theoretical state prediction is a predicted value calculated by the lightweight hydropower unit health model based on the input operating condition characteristics and the learned mapping relationship. It represents the theoretically expected value of the state monitoring quantity of the corresponding main functional chain component under the current operating conditions, in a healthy and intact state.

[0040] In this embodiment, the degree of deterioration refers to a quantitative index calculated by comparing the difference between the actual value of the condition monitoring quantity and the theoretical condition prediction quantity output by the health model of the lightweight hydropower unit.

[0041] In this embodiment, the hydropower unit deterioration trend sequence refers to the final generated time series representing the comprehensive changes in the health status of the entire hydropower unit during the assessment period. It is obtained by weighting and fusing the deterioration degree sequences of each main functional chain component within this period according to the comprehensive weights determined by the CRITIC method. This sequence serves as the direct input data for subsequent deterioration trend prediction.

[0042] In this embodiment, the input to the degradation trend prediction model is the degradation trend sequence of the hydropower unit, and the output is the future degradation trend. It consists of multiple sequence feature learning sub-models, which are weighted combinations of GRU sub-models, and the degradation trend prediction model is a Bagging-GRU model.

[0043] In this embodiment, the real-time reconstruction result refers to a set of continuous samples obtained by dividing the sequence of hydropower unit deterioration trends to be predicted into a sliding window according to the number of prediction steps. Each sample is a short sequence with a length equal to the number of prediction steps. For example, if the original sequence is [1, 2, 3, 4, 5, 6] and the number of steps is 3, then the reconstructed sample is [[1, 2, 3], [2, 3, 4], [3, 4, 5], [4, 5, 6]].

[0044] In this embodiment, such as Figure 2 As shown, the operating condition parameter extraction network refers to a complete deep autoencoder composed of the aforementioned initial encoder and initial decoder. The operating condition parameter extraction network removes interference and redundancy components from numerous operating condition parameters, reduces key information, reduces the input data scale of subsequent models, and improves the overall lightweight level.

[0045] In this embodiment, the future deterioration trend refers to the final output of the deterioration trend prediction model, that is, the sequence of predicted values ​​at multiple consecutive time points in the future, which represents the expected change trajectory of the overall health status of the hydropower unit in the future period.

[0046] In this embodiment, a degradation trend sequence is input, and a preset function is used to calculate the number of prediction steps and reconstruct the sequence to obtain real-time samples. The first sample is weighted and averaged by multiple sequence feature learning sub-models to output the first step prediction value. Based on this prediction value, subsequent input samples are reconstructed, and the weighted prediction process is repeated to obtain the future degradation trend.

[0047] In this embodiment, based on the equipment register, the unit was in its initial commissioning phase and operating well for a certain period. The operating parameters of the hydropower unit during this phase, along with the status monitoring quantities of various main functional chain components, were selected to represent its health status. Six channels of hydropower unit operating parameters were collected during this phase, including active power, head, guide vane opening, excitation current, excitation voltage, and stator coil temperature. Based on the equipment register, inspection locations, and status monitoring technical guidelines, this embodiment selected four main functional chain components affecting the safe and stable operation of the unit: the X-axis runout of the upper guide bearing, the Y-axis runout of the upper guide bearing, the X-axis runout of the lower guide bearing, and the Y-axis runout of the lower guide bearing. The runout status monitoring quantities of each component were collected. Data cleaning was completed through abnormal monitoring value removal and missing value interpolation, and a deep autoencoder was constructed, with the structure as shown below. Figure 2 As shown in the diagram. Based on the trained deep autoencoder, the reconstruction error of the deep autoencoder is minimized using MSE as the loss function. After the model converges, the result is extracted. Figure 2 The encoder part, as a pre-processing step for degradation trend assessment, is used to reduce the operating parameters of the hydropower unit. The operating parameters of the hydropower unit are input into the encoder part of the operating parameter extraction network to obtain 3D key information of the operating parameters. Based on a lightweight gradient booster, a lightweight hydropower unit health model with four main functional chain components is constructed. 90% of the data is randomly selected for training the lightweight hydropower unit health model, and subsequent data is used to verify the effectiveness of the lightweight hydropower unit health model, completing the training of four lightweight hydropower unit health models. To illustrate the advantages of the lightweight hydropower unit health model proposed in this embodiment, the proposed lightweight model is compared with health models established by artificial neural networks (ANN), Gaussian process regression (GPR), etc. The differences in various index parameters are shown in Table 1 below.

[0048] In this embodiment, it is evident that the proposed lightweight hydropower unit health model has the shortest training time among all health models. The training time for a single model is improved by an average of 384% compared to GPR and by an average of 1389% compared to ANN, saving a significant amount of training time. Furthermore, the model exhibits the best fitting error across all components. Given the limited on-site computing resources, numerous main functional chain components of the unit, and increasingly stringent monitoring requirements for each component, the selection of the lightweight hydropower unit health model not only enables rapid response, reduces system resource consumption, and lowers the frequency of system anomalies, but also achieves high-precision fitting performance, ensuring the reliability of subsequent degradation trend assessments, as shown in Table 1.

[0049] Table 1. Fitting performance of the health model of lightweight hydropower unit under different main functional chain parts.

[0050] The working principle and beneficial effects of this embodiment are as follows: Operating parameters and status monitoring data of the hydropower unit are collected; operating feature information is extracted through an operating parameter extraction network; theoretical state predictions are obtained by inputting the lightweight hydropower unit health model of each main functional chain component of the hydropower unit; the difference between the status monitoring data and the theoretical state predictions is calculated; the degradation degree of each main functional chain component of the hydropower unit is generated; and a degradation trend sequence of the hydropower unit is constructed. Finally, the degradation trend prediction model is used to reconstruct the degradation trend sequence of the hydropower unit and predict future degradation trends, thereby achieving comprehensive degradation assessment of multiple components. This significantly reduces the computational resource consumption and storage requirements of the model, improves the update efficiency of the health model and prediction model, enhances the stability of system operation, and ensures the long-term accuracy of online monitoring and prediction.

[0051] Example 2: The method provided in this embodiment collects the operating parameters of the hydropower unit and the state monitoring quantities of each main functional chain component of the hydropower unit. The operating parameters are input into a pre-trained operating parameter extraction network. Dimensionality reduction is performed on the operating parameters to obtain operating feature information. This operating feature information is then input into a pre-defined lightweight hydropower unit health model for each main functional chain component of the hydropower unit to obtain the theoretical state prediction quantities for each main functional chain component of the hydropower unit during the evaluation period, including: Collect the operating parameters of the hydropower unit and the status monitoring data of each main functional chain component of the hydropower unit during the period to be evaluated. The hydropower unit operating parameters for the period to be evaluated are input into a pre-trained operating parameter extraction network. The operating parameter extraction network performs dimensionality reduction on the hydropower unit operating parameters to determine the operating characteristic information for the period to be evaluated. The operating condition characteristic information is input into a preset lightweight hydropower unit health model. Based on the mapping relationship between the hydropower unit operating condition parameters and corresponding state monitoring quantities of each main functional chain part of the hydropower unit in the lightweight hydropower unit health model, the theoretical state prediction quantities of each main functional chain part of the hydropower unit during the period to be evaluated are output.

[0052] In this embodiment, the mapping relationship refers to the mathematical function relationship between the operating condition feature information and the theoretical state prediction quantity learned by the lightweight hydropower unit health model from historical health data during the training phase.

[0053] In this embodiment, a lightweight hydropower unit health model is initialized based on the Lightweight Gradient Boosting Machine (LightGBM) model architecture. To balance model lightweighting and accuracy, The maximum depth d of the training tree should not exceed 6, and the maximum number of leaf nodes should not exceed [a certain value]. The model does not perform random feature selection.

[0054] In this embodiment, for the i-th main functional chain part, the health status and operating condition characteristics information are used. for Input, corresponding part status monitoring quantity For the model output, train the following mapping relationship: .

[0055] The working principle and beneficial effects of this embodiment are as follows: the operating parameters and status monitoring quantities of the hydropower unit are collected, the operating feature information is obtained by dimensionality reduction through a pre-trained operating parameter extraction network, the input is input into the lightweight hydropower unit health model, and the theoretical status prediction quantities of each main functional chain part are output according to the preset mapping relationship, so as to realize the comprehensive health assessment of multiple parts, reduce the dimensionality of input data and the consumption of computing resources, improve the output efficiency of theoretical status prediction quantities and the model update speed, and support long-term accurate monitoring of deterioration trends.

[0056] Example 3: The method provided in this embodiment of the present disclosure, wherein the preset operating condition parameter extraction network includes: Construct an initial encoder with a preset number of layers, decreasing the number of neurons in each layer from the shallowest to the deepest, set an activation function for each layer, and transpose the parameters of each layer of the initial encoder to obtain an initial decoder with the same structure as the initial encoder. The initial decoder and the initial encoder are combined to obtain the operating condition parameter extraction network.

[0057] In this embodiment, the preset number of layers refers to the total number of network layers that are manually set before the neural network is constructed.

[0058] In this embodiment, the initial encoder refers to the part of the neural network first constructed during the construction of the deep autoencoder, which is used for compression and feature extraction of the input data. Its structural characteristic is that the number of neurons in each layer decreases progressively from the shallowest to the deepest.

[0059] In this embodiment, the activation function refers to the function set for each layer of the neural network to introduce nonlinear transformations.

[0060] In this embodiment, parameter transpose refers to a symmetric method for constructing a decoder. After obtaining the trained initial encoder, the weight matrix of each layer is transposed and reversed to quickly obtain an initial decoder that is symmetric to the encoder structure.

[0061] In this embodiment, the initial decoder refers to the part of the neural network that is symmetrical to the initial encoder structure, constructed through methods such as parameter transposition. It decodes the low-dimensional feature representation obtained from the encoder back into the original high-dimensional data space.

[0062] The working principle and beneficial effects of this embodiment are as follows: Input data is compressed by an encoder with progressively decreasing neurons, its parameters are transposed to construct a symmetric decoder, and combined to form an autoencoder network that can reconstruct the input, thereby achieving automatic dimensionality reduction and feature extraction of data and generating low-dimensional key information for use by downstream models.

[0063] Example 4: The method provided in this embodiment compares the difference between the monitored state quantity and the theoretical state prediction quantity based on the mapping relationship, generates the degree of deterioration of each main functional chain component of the hydropower unit during the evaluation period, analyzes the degree of deterioration of each main functional chain component of the hydropower unit, and generates a deterioration trend sequence of the hydropower unit during the evaluation period based on the analysis results, including: Based on the mapping relationship, the difference between the state monitoring quantity and the theoretical state prediction quantity of each main functional chain part of the hydropower unit is compared, and the deterioration degree of each main functional chain part of the hydropower unit to be evaluated during the corresponding evaluation period is generated. The standard deviation is used to measure the variability of the deterioration degree of each main functional chain component of the hydropower unit during the evaluation period, and the correlation coefficient is used to measure the conflict of the deterioration degree of each main functional chain component of the hydropower unit during the evaluation period. The comprehensive weights of each main functional chain component of the hydropower unit are calculated based on the variability and conflict, and the deterioration trend sequence of the hydropower unit for the evaluation period is generated by combining the CRITIC method with the comprehensive weights.

[0064] In this embodiment, variability refers to the magnitude or dispersion of the degradation sequence of the main functional chain component being evaluated during the evaluation period. High variability means that the degradation state of that component is unstable and fluctuates drastically, potentially containing more information that requires attention.

[0065] In this embodiment, conflict refers to the degree of opposition or contradiction in the degradation sequence changes between different main functional chain parts, measured by the correlation coefficient. If the degradation trends of two parts are highly negatively correlated, their conflict is high, indicating that there is contradictory or competing information, which requires careful consideration during comprehensive evaluation.

[0066] In this embodiment, the process of generating the day-k degradation degree of each main functional chain component of the unit involves sending the day-k operating condition characteristic information into the health model of each lightweight hydropower unit. The theoretical state monitoring quantities for each main functional chain component are obtained: ,in, This represents the theoretical state monitoring quantity of the i-th main functional chain part on day k. This represents the operating condition characteristics on day k; N represents the total number of main functional chain components; the degradation degree of each main functional chain component on day k is calculated as follows: ,in, This represents the degradation degree of the i-th main function chain part on day k; This represents the status monitoring data of the i-th part on day k; This represents the theoretical state monitoring value of the i-th part on day k; This indicates that the i-th main function chain part is in the first position. Number of samples per day The value is a small positive number close to 0 to prevent the denominator from being zero; K represents the total length of the period to be evaluated; the degradation degree of each main functional chain component of the unit during the period to be evaluated is obtained. By combining the CRITIC evaluation method to assess the weight of the deterioration degree of each main functional chain component, the deterioration trend of the unit can be obtained.

[0067] In this embodiment, the formula for calculating variability is as follows: ,in, This represents the variability of the i-th main functional chain part; This represents the time-interval average of the degradation degree of the i-th main functional chain component.

[0068] In this embodiment, the calculation formula corresponding to the conflict is as follows: ,in, This indicates the degradation degree of the j-th main function chain component; This represents the degree of conflict between the degradation degree of the i-th main function chain component and the degradation degree of the j-th main function chain component. , They represent the first The, the The average degradation degree of each main functional chain component over a given period. The overall weight corresponding to each main functional chain component is calculated based on variability and conflict. , This represents the comprehensive weight corresponding to the i-th main functional chain component. The comprehensive weight obtained using the CRITIC method is then used to generate a hydropower unit deterioration trend sequence. : .

[0069] In this embodiment, the comprehensive weight refers to a combined weight calculated for the degradation degree of each main functional chain component. It considers the variability of the degradation degree of the component itself and its conflict with the degradation degrees of other components. The comprehensive weight is used to determine the proportion of degradation degree of each component when synthesizing the overall trend.

[0070] In this embodiment, the CRITIC method is an objective multi-index weight determination method. It uses the degree of deterioration of each part as an index, its standard deviation to measure variability, and its correlation coefficient to measure conflict, thereby calculating the comprehensive weight of each part.

[0071] The working principle and beneficial effects of this embodiment are as follows: the degree of degradation is obtained by calculating the difference between the condition monitoring quantity and the theoretical condition prediction quantity. The variation and conflict are quantified by standard deviation and correlation coefficient. The comprehensive degradation trend sequence is generated by weighting the CRITIC method, so as to achieve objective integration of the degradation degree of multiple parts and improve the comprehensiveness and accuracy of the overall degradation trend assessment of the unit.

[0072] Example 5: The method provided in this embodiment of the present disclosure inputs the hydropower unit deterioration trend sequence into a preset deterioration trend prediction model, reconstructs the hydropower unit deterioration trend sequence, and predicts the future deterioration trend of the hydropower unit based on the real-time reconstruction results, including: The hydropower unit deterioration trend sequence is input into a preset deterioration trend prediction model. A preset function is used to determine the prediction correlation steps of the hydropower unit deterioration trend sequence. The hydropower unit deterioration trend sequence is then reconstructed based on the prediction correlation steps to obtain real-time reconstruction results. Based on the first sample input in the real-time reconstruction result, several sequence feature learning sub-models in the degradation trend prediction model are used to perform a weighted average of several prediction results according to the sub-model weights corresponding to the several sequence feature learning sub-models, and output the prediction value of the first future time point. The rolling prediction of all samples in the real-time reconstruction results yields the predicted value of the corresponding sample, and the future deterioration trend of the hydropower unit is obtained based on all the predicted values.

[0073] In this embodiment, the prediction correlation step number refers to the optimal historical window length determined by analyzing the historical hydropower unit deterioration trend sequence using a preset function. It indicates how many consecutive time points in the past need to be referenced when predicting a future point; the preset function may be a partial autocorrelation function.

[0074] In this embodiment, the sequence feature learning sub-model refers to the specific base learner in the degradation trend prediction model, namely the gated recurrent unit (GRU) neural network. Each sub-model is trained independently on different subsets of data generated by the Bagging method to learn the nonlinear mapping relationship from historical sequences to future values, that is, to capture the temporal dependency features of the sequence.

[0075] In this embodiment, the predicted value refers to the prediction result of the model at the next time step after the deterioration trend sequence of the hydropower unit is input into the deterioration trend prediction model. For multi-step prediction, the first predicted value is combined with historical data to generate a new sample for predicting time t+1, and so on.

[0076] In this embodiment, as a key technology supporting condition-based maintenance of generating units, the current degradation trend assessment and prediction technology processes relatively limited information sources, requiring significant computational resources and training time for practical application, making it difficult to meet the rapid data mining needs under massive data and multiple monitoring channels. Against this backdrop, this disclosure proposes a lightweight hydropower unit degradation trend assessment and prediction method, balancing accuracy and model lightweighting. It utilizes a deep autoencoder (operating condition parameter extraction network) to remove interference and redundancy components from operating condition parameters, reducing the input data scale of subsequent models and improving lightweighting from a data perspective. Furthermore, it constructs a lightweight LightGBM health model and a low-parameter Bagging-GRU prediction model to improve model response speed under massive measurement points, ensuring rapid and reliable degradation assessment and prediction from the model level. Based on the above aspects, the lightweight hydropower unit degradation trend assessment and prediction method provided by this disclosure is scientific and practical, and has significant value for the rapid deployment and updating of hydropower unit degradation assessment and prediction methods.

[0077] The working principle and beneficial effects of this embodiment are as follows: An input degradation trend sequence is used, and a preset function is employed to calculate the number of prediction steps and reconstruct the sequence to obtain real-time samples. The first sample is weighted and averaged by multiple sequence feature learning sub-models to output the first-step prediction value. Based on this prediction value, subsequent input samples are reconstructed on a rolling basis, and the weighted prediction process is repeated to obtain multi-step future degradation trends, achieving multi-step rolling integrated prediction and improving the stability and accuracy of the prediction results.

[0078] Example 6: The method provided in this embodiment of the present disclosure, such as... Figure 5 As shown, the preset degradation trend prediction model includes: Obtain historical deterioration trend sequences of hydropower units; The relevant number of steps of the historical hydropower unit deterioration trend sequence is determined by a preset function, the historical hydropower unit deterioration trend sequence is reconstructed according to the relevant number of steps, and sample-label pairs are generated according to the reconstruction results. The sample-label pairs are divided into training set and test set based on a preset percentage. The training set is resampled several times with replacement using the Bagging method to obtain several sub-training sample sets with the same number of samples as the training set. Sequence feature learning sub-models are constructed for each sub-training sample set and the models are trained. Calculate the mean squared error of each sequence feature learning sub-model on the corresponding sub-training sample set, and determine the sub-model weight of each sequence feature learning sub-model based on the mean squared error. Then, perform weighted integration of each sequence feature learning sub-model based on the sub-model weight to obtain the degradation trend prediction model. The accuracy of the degradation trend prediction model is verified using the test set. If the preset accuracy requirement is met, the degradation trend prediction model is output.

[0079] In this embodiment, the historical hydropower unit deterioration trend sequence refers to the time series data generated over a period of time during the model training phase, which represents the overall changes in the health status of the hydropower units.

[0080] In this embodiment, the sample-label pair refers to the supervised learning data formed by reconstructing the historical hydropower unit deterioration trend sequence according to the relevant number of steps. A sample is a continuous historical data subsequence of length equal to the number of relevant steps. The label is the actual observation value at the next time point immediately following this subsequence.

[0081] In this embodiment, the preset percentage refers to a pre-set ratio when dividing the training set and the test set. This ratio is used to divide all sample-label pairs, with one portion used for training the model and the other for testing model performance.

[0082] In this embodiment, resampling with replacement refers to randomly selecting a sample from the training set and then putting that sample back into the original training set, so that it may still be selected in the next sampling. By repeating this process, multiple new datasets can be generated.

[0083] In this embodiment, a sub-training sample set refers to multiple datasets generated through resampling with replacement, which have the same number of samples as the original training set but different sample compositions. Each sub-training sample set is used to independently train a sequence feature learning sub-model.

[0084] In this embodiment, the input to the sequence feature learning sub-model is samples from the sub-training sample set, and the output is the predicted value of the label. The sequence feature learning sub-model is... Figure 5 The GRU sub-model in the model.

[0085] In this embodiment, weighted ensemble refers to a method of combining multiple trained sequence feature learning sub-models into a stronger model. The prediction results of each GRU sub-model (sequence feature learning sub-model) are weighted and averaged according to their respective sub-model weights to obtain the final ensemble prediction result.

[0086] In this embodiment, the performance of each GRU sub-model on the corresponding sub-training set is calculated. : ,in, 'a' represents the number of samples in the sub-training set; 'a' represents the sub-model index; 'b' represents the sample index. The degree of degradation predicted by the GRU sub-model for sample b; To calculate the true degradation degree of sample b in the GRU sub-model, calculate the sub-model weights of each GRU sub-model. : , where M represents the total number of GRU sub-models.

[0087] In this embodiment, the preset accuracy requirement refers to a pre-set performance threshold used to determine whether the degradation trend prediction model is qualified. It is typically measured by calculating the prediction error on a test set; if the error is below this threshold, the model meets the requirements and can be used.

[0088] In this embodiment, the partial autocorrelation function (PACF) is used to determine the relevant steps of the unit's deterioration trend, and the deterioration trend sequence is reconstructed accordingly to generate sample-label pairs. The first few steps of each sample-label pair are then... The training set is used as the training set, and the remaining portion is used as the test set for model testing. The training set is resampled M times with replacement using the Bagging method to obtain M sub-training sets with the same number of samples as the training set. A GRU model is built for each sub-training set and trained. The MSE of each GRU model on the corresponding sub-training set is calculated, and the weights of each GRU model are determined accordingly. The final Bagging-GRU prediction model is obtained based on the weights. The accuracy of the Bagging-GRU model is verified using the test set. If the accuracy requirements are met, it can be used for degradation trend prediction.

[0089] In this embodiment, based on the CRITIC evaluation method, the degradation trends of each main functional chain component are integrated. The weights of the degradation trends of the upper guide bearing's X-direction swing, the upper guide bearing's Y-direction swing, the lower guide bearing's X-direction swing, and the lower guide bearing's Y-direction swing are 0.236, 0.288, 0.248, and 0.228, respectively. The unit degradation trend is then generated accordingly. Figure 4 As shown in the figure, the unit's degradation level exhibits a fluctuating upward trend with increasing operating time, consistent with the feedback from the on-site operation and maintenance report. This demonstrates that the degradation trend assessment method proposed in this invention can effectively assess the unit's degradation status. Based on the Bagging method, a lightweight GRU sub-model with few parameters is constructed, as shown in the figure. Figure 5 The Bagging-GRU prediction model is shown. By analyzing the autocorrelation of the unit deterioration trend sequence using the partial autocorrelation function, it can be seen that the partial autocorrelation function falls within the 95% confidence interval for the first 6 delay steps. Therefore, the number of correlation steps is determined to be 6, and the deterioration trend sequence is reconstructed accordingly, generating a total of 116 sample-label pairs. The first 86 sample-label pairs are used as the training set, and the last 30 sample-label pairs are used as the test set. The Bagging method is then used to process the training set... Second resampling, to obtain For each sub-training sample set, a GRU sub-model is constructed and trained. The training results show that when… When the model training effect is optimal, the Bagging-GRU prediction model (degradation trend prediction model) is compared with the GRU sub-model and the ordinary GRU model. The parameters and prediction performance of each model are shown in Table 2.

[0090] Table 2. Parameter Quantity and Prediction Performance of GRU Sub-models, Bagging-GRU, and Ordinary GRU

[0091] In this embodiment, such as Figure 3 As shown, under conditions of no overfitting and sufficient training, the individual GRU sub-models constituting Bagging-GRU have the smallest parameter size and the worst prediction performance. In contrast, the parameter size of ordinary GRU is 1.64 times that of Bagging-GRU, but its performance in RMSE, MAPE, MAE, R2, and other metrics is inferior to Bagging-GRU. This confirms that the Bagging method can improve the prediction performance of GRU models, achieving optimal prediction performance with small parameter size, low computational resources, and short training time. The trend of prediction degradation is as follows: Figure 6 As shown, by Figure 6 As can be seen, the predicted degradation trend can accurately reflect the degradation evolution of the unit. Therefore, the Bagging-GRU lightweight prediction model proposed in this embodiment performs well, balancing model accuracy and lightweight design. The greater the number of channels and the amount of data, the more obvious the advantages of this embodiment become. It is suitable for scenarios with multiple channels, frequent model updates, and high real-time requirements, such as centralized control centers.

[0092] The working principle and beneficial effects of this embodiment are as follows: Based on the historical hydropower unit deterioration trend sequence, the relevant number of steps is determined, sample-label pairs are reconstructed, and the dataset is divided. Multiple sequence feature learning sub-models are trained on the training sample set through Bagging resampling. The weights are determined according to the mean square error of each sequence feature learning sub-model, and they are weighted and integrated. Finally, the output prediction model is verified. Through ensemble learning, the model variance and overfitting risk are reduced, and the generalization ability and robustness of the prediction model are improved.

[0093] To achieve the above embodiments, this disclosure also proposes a lightweight hydropower unit deterioration trend assessment and prediction device.

[0094] Figure 7 This is a schematic diagram of the lightweight hydropower unit degradation trend assessment and prediction device provided in an embodiment of this disclosure. The device 200 can be implemented by software and / or hardware, and is generally integrated into an electronic device. Figure 7 As shown, the device 200 includes: a data acquisition module 201, an analysis module 202, and a prediction module 203, wherein, The data acquisition module 201 is used to acquire the operating parameters of the hydropower unit and the status monitoring quantities of each main functional chain component of the hydropower unit. The operating parameters of the hydropower unit are input into a pre-trained operating parameter extraction network, and the operating parameters of the hydropower unit are dimensionality reduced to obtain operating feature information. The operating feature information is input into a preset lightweight hydropower unit health model for each main functional chain component of the hydropower unit to obtain the theoretical state prediction quantities of each main functional chain component of the hydropower unit during the period to be evaluated. The lightweight hydropower unit health model is constructed based on a lightweight gradient boosting mechanism. Analysis module 202 is used to compare the difference between the state monitoring quantity and the theoretical state prediction quantity based on the mapping relationship, generate the deterioration degree of each main functional chain part of the hydropower unit during the period to be evaluated, analyze the deterioration degree of each main functional chain part of the hydropower unit, and generate the deterioration trend sequence of the hydropower unit during the period to be evaluated based on the analysis results. The prediction module 203 is used to input the deterioration trend sequence of the hydropower unit into a preset deterioration trend prediction model, reconstruct the deterioration trend sequence of the hydropower unit, and predict the future deterioration trend of the hydropower unit based on the real-time reconstruction results.

[0095] The device provided in this embodiment of the disclosure, wherein the acquisition module 201 is specifically used for: Collect the operating parameters of the hydropower unit and the status monitoring data of each main functional chain component of the hydropower unit during the period to be evaluated. The hydropower unit operating parameters for the period to be evaluated are input into a pre-trained operating parameter extraction network. The operating parameter extraction network performs dimensionality reduction on the hydropower unit operating parameters to determine the operating characteristic information for the period to be evaluated. The operating condition characteristic information is input into a preset lightweight hydropower unit health model. Based on the mapping relationship between the hydropower unit operating condition parameters and corresponding state monitoring quantities of each main functional chain part of the hydropower unit in the lightweight hydropower unit health model, the theoretical state prediction quantities of each main functional chain part of the hydropower unit during the period to be evaluated are output.

[0096] The device provided in this embodiment of the disclosure, wherein the acquisition module 201 is specifically used for: Construct an initial encoder with a preset number of layers, decreasing the number of neurons in each layer from the shallowest to the deepest, set an activation function for each layer, and transpose the parameters of each layer of the initial encoder to obtain an initial decoder with the same structure as the initial encoder. The initial decoder and the initial encoder are combined to obtain the operating condition parameter extraction network.

[0097] The apparatus provided in this disclosure, wherein the analysis module 202 is specifically used for: Based on the mapping relationship, the difference between the state monitoring quantity and the theoretical state prediction quantity of each main functional chain part of the hydropower unit is compared, and the deterioration degree of each main functional chain part of the hydropower unit to be evaluated during the corresponding evaluation period is generated. The standard deviation is used to measure the variability of the deterioration degree of each main functional chain component of the hydropower unit during the evaluation period, and the correlation coefficient is used to measure the conflict of the deterioration degree of each main functional chain component of the hydropower unit during the evaluation period. The comprehensive weights of each main functional chain component of the hydropower unit are calculated based on the variability and conflict, and the deterioration trend sequence of the hydropower unit for the evaluation period is generated by combining the CRITIC method with the comprehensive weights.

[0098] The apparatus provided in this disclosure, wherein the prediction module 203 is specifically used for: The hydropower unit deterioration trend sequence is input into a preset deterioration trend prediction model. A preset function is used to determine the prediction correlation steps of the hydropower unit deterioration trend sequence. The hydropower unit deterioration trend sequence is then reconstructed based on the prediction correlation steps to obtain real-time reconstruction results. Based on the first sample input in the real-time reconstruction result, several sequence feature learning sub-models in the degradation trend prediction model are used to perform a weighted average of several prediction results according to the sub-model weights corresponding to the several sequence feature learning sub-models, and output the prediction value of the first future time point. The rolling prediction of all samples in the real-time reconstruction results yields the predicted value of the corresponding sample, and the future deterioration trend of the hydropower unit is obtained based on all the predicted values.

[0099] The apparatus provided in this disclosure, wherein the prediction module 203 is specifically used for: Obtain historical deterioration trend sequences of hydropower units; The relevant number of steps of the historical hydropower unit deterioration trend sequence is determined by a preset function, the historical hydropower unit deterioration trend sequence is reconstructed according to the relevant number of steps, and sample-label pairs are generated according to the reconstruction results. The sample-label pairs are divided into training set and test set based on a preset percentage. The training set is resampled several times with replacement using the Bagging method to obtain several sub-training sample sets with the same number of samples as the training set. Sequence feature learning sub-models are constructed for each sub-training sample set and the models are trained. Calculate the mean squared error of each sequence feature learning sub-model on the corresponding sub-training sample set, and determine the sub-model weight of each sequence feature learning sub-model based on the mean squared error. Then, perform weighted integration of each sequence feature learning sub-model based on the sub-model weight to obtain the degradation trend prediction model. The degradation trend prediction was verified using the test set.

[0100] The lightweight hydropower unit degradation trend assessment and prediction device provided in this disclosure can execute the lightweight hydropower unit degradation trend assessment and prediction method provided in any embodiment of this disclosure, and has the corresponding functional modules and beneficial effects of the execution method.

[0101] To implement the above embodiments, this disclosure also proposes a computer program product, including a computer program / instruction, which, when executed by a processor, implements the lightweight hydropower unit deterioration trend assessment and prediction method in the above embodiments.

[0102] Figure 8 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this disclosure.

[0103] The following is a detailed reference. Figure 8 The diagram illustrates a structural schematic suitable for implementing the electronic device 300 in the embodiments of this disclosure. The electronic device 300 in the embodiments of this disclosure may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), in-vehicle terminals (e.g., in-vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. Figure 8 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments disclosed herein.

[0104] like Figure 8 As shown, the electronic device 300 may include a processor (e.g., a central processing unit, a graphics processing unit, etc.) 301, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 302 or a program loaded from a memory 308 into a random access memory (RAM) 303. The RAM 303 also stores various programs and data required for the operation of the electronic device 300. The processor 301, ROM 302, and RAM 303 are interconnected via a bus 304. An input / output (I / O) interface 305 is also connected to the bus 304.

[0105] Typically, the following devices can be connected to I / O interface 305: input devices 306 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 307 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; memory devices 308 including, for example, magnetic tapes, hard disks, etc.; and communication devices 309. Communication device 309 allows electronic device 300 to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 8An electronic device 300 with various devices is shown; however, it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed alternatively.

[0106] In particular, according to embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this disclosure include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device 309, or installed from a memory 308, or installed from a ROM 302. When the computer program is executed by the processor 301, it performs the functions defined in the lightweight hydropower unit deterioration trend assessment and prediction method of embodiments of this disclosure.

[0107] It should be noted that the computer-readable medium described in this disclosure can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this disclosure, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this disclosure, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0108] In some implementations, clients and servers can communicate using any currently known or future-developed network protocol such as HTTP (Hypertext Transfer Protocol) and can interconnect with digital data communication (e.g., communication networks) of any form or medium. Examples of communication networks include local area networks (“LANs”), wide area networks (“WANs”), the Internet (e.g., the Internet of Things), and peer-to-peer networks (e.g., ad hoc peer-to-peer networks), as well as any currently known or future-developed networks.

[0109] The aforementioned computer-readable medium may be included in the aforementioned electronic device; or it may exist independently and not assembled into the electronic device.

[0110] The aforementioned computer-readable medium carries one or more programs, which, when executed by the electronic device, cause the electronic device to perform the aforementioned method for assessing and predicting the deterioration trend of lightweight hydropower units.

[0111] Electronic devices can be programmed with computer program code in one or more programming languages ​​or combinations thereof to perform the operations of this disclosure. These programming languages ​​include, but are not limited to, object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as "C" or similar languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0112] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0113] The units described in the embodiments of this disclosure can be implemented in software or hardware. The names of the units are not, in some cases, intended to limit the specific unit.

[0114] The functions described above in this document can be performed at least in part by one or more hardware logic components. For example, exemplary types of hardware logic components that can be used, without limitation, include: field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), system-on-a-chip (SoCs), complex programmable logic devices (CPLDs), and so on.

[0115] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0116] The above description is merely a preferred embodiment of this disclosure and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of this disclosure is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features disclosed in this disclosure that have similar functions.

[0117] Furthermore, while the operations are described in a specific order, this should not be construed as requiring these operations to be performed in the specific order shown or in a sequential order. In certain environments, multitasking and parallel processing may be advantageous. Similarly, while several specific implementation details are included in the above discussion, these should not be construed as limiting the scope of this disclosure. Certain features described in the context of individual embodiments may also be implemented in combination in a single embodiment. Conversely, various features described in the context of a single embodiment may also be implemented individually or in any suitable sub-combination in multiple embodiments.

[0118] Although the subject matter has been described using language specific to structural features and / or methodological logic, it should be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or actions described above. Rather, the specific features and actions described above are merely illustrative examples of implementing the claims.

Claims

1. A method for predicting the deterioration trend of lightweight hydropower units, characterized in that, include: The operating parameters of the hydropower unit and the status monitoring quantities of each main functional chain component of the hydropower unit are collected. The operating parameters of the hydropower unit are input into a pre-trained operating parameter extraction network. The operating parameters of the hydropower unit are dimensionality reduced to obtain operating feature information. The operating feature information is input into a preset lightweight hydropower unit health model for each main functional chain component of the hydropower unit to obtain the theoretical state prediction quantity of each main functional chain component of the hydropower unit during the evaluation period. The lightweight hydropower unit health model is based on a lightweight gradient boosting mechanism. Based on the mapping relationship, the difference between the state monitoring quantity and the theoretical state prediction quantity is compared, the deterioration degree of each main functional chain part of the hydropower unit during the evaluation period is generated, the deterioration degree of each main functional chain part of the hydropower unit is analyzed, and the deterioration trend sequence of the hydropower unit during the evaluation period is generated based on the analysis results. The deterioration trend sequence of the hydropower unit is input into a preset deterioration trend prediction model to reconstruct the deterioration trend sequence of the hydropower unit, and the future deterioration trend of the hydropower unit is predicted based on the real-time reconstruction results.

2. The method according to claim 1, characterized in that, The system collects operating parameters of the hydropower unit and status monitoring data of each main functional chain component. The operating parameters are input into a pre-trained operating parameter extraction network. Dimensionality reduction is performed on the operating parameters to obtain operating feature information. This feature information is then input into a pre-defined lightweight hydropower unit health model for each main functional chain component to obtain theoretical state predictions for each main functional chain component during the evaluation period. These predictions include: Collect the operating parameters of the hydropower unit and the status monitoring data of each main functional chain component of the hydropower unit during the period to be evaluated. The hydropower unit operating parameters for the period to be evaluated are input into a pre-trained operating parameter extraction network. The operating parameter extraction network performs dimensionality reduction on the hydropower unit operating parameters to determine the operating characteristic information for the period to be evaluated. The operating condition characteristic information is input into a preset lightweight hydropower unit health model. Based on the mapping relationship between the hydropower unit operating condition parameters and corresponding state monitoring quantities of each main functional chain part of the hydropower unit in the lightweight hydropower unit health model, the theoretical state prediction quantities of each main functional chain part of the hydropower unit during the period to be evaluated are output.

3. The method according to claim 2, characterized in that, The preset operating condition parameter extraction network includes: Construct an initial encoder with a preset number of layers, decreasing the number of neurons in each layer from the shallowest to the deepest, set an activation function for each layer, and transpose the parameters of each layer of the initial encoder to obtain an initial decoder with the same structure as the initial encoder. The initial decoder and the initial encoder are combined to obtain the operating condition parameter extraction network.

4. The method according to claim 1, characterized in that, Based on the mapping relationship, the difference between the monitored state quantities and the theoretical state prediction quantities is compared to generate the deterioration degree of each main functional chain component of the hydropower unit during the evaluation period. The deterioration degree of each main functional chain component of the hydropower unit is analyzed, and a deterioration trend sequence of the hydropower unit during the evaluation period is generated based on the analysis results, including: Based on the mapping relationship, the difference between the state monitoring quantity and the theoretical state prediction quantity of each main functional chain part of the hydropower unit is compared, and the deterioration degree of each main functional chain part of the hydropower unit to be evaluated during the corresponding evaluation period is generated. The standard deviation is used to measure the variability of the deterioration degree of each main functional chain component of the hydropower unit during the evaluation period, and the correlation coefficient is used to measure the conflict of the deterioration degree of each main functional chain component of the hydropower unit during the evaluation period. The comprehensive weights of each main functional chain component of the hydropower unit are calculated based on the variability and conflict, and the deterioration trend sequence of the hydropower unit for the evaluation period is generated by combining the CRITIC method with the comprehensive weights.

5. The method according to claim 1, characterized in that, The degradation trend sequence of the hydropower unit is input into a preset degradation trend prediction model to reconstruct the degradation trend sequence of the hydropower unit, and the future degradation trend of the hydropower unit is predicted based on the real-time reconstruction results, including: The hydropower unit deterioration trend sequence is input into a preset deterioration trend prediction model. A preset function is used to determine the prediction correlation steps of the hydropower unit deterioration trend sequence. The hydropower unit deterioration trend sequence is then reconstructed based on the prediction correlation steps to obtain real-time reconstruction results. Based on the first sample input in the real-time reconstruction result, several sequence feature learning sub-models in the degradation trend prediction model are used to perform a weighted average of several prediction results according to the sub-model weights corresponding to the several sequence feature learning sub-models, and output the prediction value of the first future time point. The rolling prediction of all samples in the real-time reconstruction results yields the predicted value of the corresponding sample, and the future deterioration trend of the hydropower unit is obtained based on all the predicted values.

6. The method according to claim 5, characterized in that, The preset degradation trend prediction model includes: Obtain historical deterioration trend sequences of hydropower units; The relevant number of steps of the historical hydropower unit deterioration trend sequence is determined by a preset function, the historical hydropower unit deterioration trend sequence is reconstructed according to the relevant number of steps, and sample-label pairs are generated according to the reconstruction results. The sample-label pairs are divided into training set and test set based on a preset percentage. The training set is resampled several times with replacement using the Bagging method to obtain several sub-training sample sets with the same number of samples as the training set. Sequence feature learning sub-models are constructed for each sub-training sample set and the models are trained. Calculate the mean squared error of each sequence feature learning sub-model on the corresponding sub-training sample set, and determine the sub-model weight of each sequence feature learning sub-model based on the mean squared error. Then, perform weighted integration of each sequence feature learning sub-model based on the sub-model weight to obtain the degradation trend prediction model. The accuracy of the degradation trend prediction model is verified using the test set. If the preset accuracy requirement is met, the degradation trend prediction model is output.

7. A device for assessing and predicting the deterioration trend of a lightweight hydropower unit, the device comprising: The data acquisition module is used to collect the operating parameters of the hydropower unit and the status monitoring data of each main functional chain component of the hydropower unit. The operating parameters of the hydropower unit are input into a pre-trained operating parameter extraction network, and the operating parameters of the hydropower unit are dimensionality reduced to obtain operating feature information. The operating feature information is input into a preset lightweight hydropower unit health model for each main functional chain component of the hydropower unit to obtain the theoretical state prediction of each main functional chain component of the hydropower unit during the evaluation period. The lightweight hydropower unit health model is based on a lightweight gradient boosting mechanism. The analysis module is used to compare the difference between the state monitoring quantity and the theoretical state prediction quantity based on the mapping relationship, generate the degree of deterioration of each main functional chain part of the hydropower unit during the period to be evaluated, analyze the degree of deterioration of each main functional chain part of the hydropower unit, and generate the deterioration trend sequence of the hydropower unit during the period to be evaluated based on the analysis results. The prediction module is used to input the deterioration trend sequence of the hydropower unit into a preset deterioration trend prediction model, reconstruct the deterioration trend sequence of the hydropower unit, and predict the future deterioration trend of the hydropower unit based on the real-time reconstruction results.

8. An electronic device, characterized in that, include: Memory; processor; as well as Computer programs; The computer program is stored in the memory and configured to be executed by the processor to implement the lightweight hydropower unit deterioration trend assessment and prediction method as described in any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, It stores a computer program / instruction thereon, which, when executed by a processor, implements the steps of the method described in any one of claims 1-6.

10. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the method according to any one of claims 1 to 6.

Images