Rapid early warning system and method for deterioration state trend of water energy storage unit

Abstract

The invention relates to the technical field of intelligent early warning of a water energy storage unit, in particular to a rapid early warning system and method for the deterioration state trend of the water energy storage unit, and the system comprises a data collection module which is used for carrying out the real-time collection of the operation data of the water energy storage unit; the feature processing module is used for performing feature processing on the operation data of the water energy storage unit; the real-time monitoring module is used for constructing a real-time monitoring model and monitoring the degradation state of the water energy storage unit in real time; the degradation prediction module is used for constructing a degradation prediction model and predicting the degradation state trend of the water energy storage unit; the prediction adjustment module is used for adjusting the degradation prediction module; and the prediction optimization module is used for optimizing the prediction adjustment module, and the prediction optimization module is connected with the prediction adjustment module. According to the invention, the operation state of the water energy storage unit can be monitored and predicted in real time, and rapid early warning is carried out on the deterioration state trend, so that the operation reliability and the maintenance efficiency are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent early warning of water storage units, and in particular to a system and method for rapid early warning of degradation state trends of water storage units. Background Art

[0002] Pumped-storage power stations play an important role in the power system, especially in the context of the rapid development of renewable energy. As the requirements for flexibility and reliability of the power grid continue to increase, pumped-storage power stations are not just simple energy storage facilities, but have also become an important part of power dispatching and frequency regulation. Therefore, the maintenance and management of pumped-storage units have become particularly important. The existing water storage unit operation status monitoring system mainly relies on regular manual inspections and simple fault alarm systems, which have problems such as poor real-time performance, low accuracy, and delayed warnings, and cannot effectively prevent failures caused by unit degradation. Therefore, there is an urgent need for a system and method that can monitor and predict the degradation status of water storage units in real time to improve the reliability of unit operation and maintenance efficiency.

[0003] Chinese Patent Publication No.: CN113268835B discloses a method and device for health assessment of pumped-storage hydroelectric units based on extended isolation forests and cloud models, including obtaining the current steady-state data of the bearing swing of the pumped-storage hydroelectric unit in real-time operation, obtaining the current abnormal score value of the pumped-storage hydroelectric unit in real-time operation according to the current steady-state data, cloud model and extended isolation forest model; normalizing the current abnormal score value using a preset abnormal score threshold range to obtain the current health of the pumped-storage hydroelectric unit in real-time operation; and using the historical health of the pumped-storage hydroelectric unit and the current health to assess the current health status of the pumped-storage hydroelectric unit. However, this solution cannot predict the operating status of the hydroelectric unit, and it is difficult to improve the operating reliability and maintenance efficiency. Summary of the invention

[0004] To this end, the present invention provides a rapid early warning system and method for the degradation state trend of a water storage unit, so as to overcome the problems of low operating reliability and maintenance efficiency of the water storage unit in the prior art.

[0005] To achieve the above object, the present invention provides a rapid early warning system for degradation trend of a water storage unit, comprising:

[0006] Data acquisition module, used to collect the operation data of the water storage unit in real time;

[0007] A feature processing module is used to perform feature processing on the operation data of the water storage unit to obtain feature operation data;

[0008] Real-time monitoring module, used to build a real-time monitoring model and monitor the degradation status of the water storage unit in real time;

[0009] Deterioration prediction module, used to build a degradation prediction model and predict the degradation state trend of the hydropower storage unit;

[0010] A prediction adjustment module, used to adjust the degradation prediction module according to the prediction results and identification results of each time node;

[0011] The prediction optimization module is used to optimize the prediction adjustment module according to the number of adjustments within the optimization cycle and the preset adjustments.

[0012] Furthermore, the feature processing module includes:

[0013] A preprocessing unit, used to preprocess the operation data of the water storage unit to obtain preprocessed operation data;

[0014] The feature extraction unit is used to construct a feature extraction model and extract features from the preprocessing operation data according to the feature extraction model to obtain feature operation data. The feature extraction unit is connected to the preprocessing unit.

[0015] Further, the preprocessing unit preprocesses the operation data of the water storage unit through a data editing tool to obtain preprocessed operation data;

[0016] The feature extraction unit, when constructing a feature extraction model, divides the historical feature extraction database into a feature training set of 75%, a feature verification set of 15%, and a feature test set of 10%, inputs the feature training set into the convolutional neural network model to train the convolutional neural network model, inputs the feature verification set into the trained convolutional neural network model to iteratively optimize the parameters of the trained convolutional neural network model, and inputs the feature test set into the iteratively optimized convolutional neural network model for testing, sets the total number of samples in the feature test set to M0, the number of samples correctly output by the iteratively optimized convolutional neural network model to M1, and the feature accuracy to Q, Q=M1 / M0, compares the feature accuracy Q with the preset feature accuracy Q0, judges the training compliance of the convolutional neural network model according to the comparison result, and outputs according to the judgment result, wherein:

[0017] When Q≥Q0, it is determined that the convolutional neural network model training has reached the standard, the convolutional neural network model is output as a feature extraction model, and the preprocessed operation data is input into the feature extraction model for feature extraction to obtain feature operation data;

[0018] When Q<Q0, it is determined that the training of the convolutional neural network model does not meet the standards, the historical feature extraction database is updated, and the convolutional neural network model is initialized. The initialized convolutional neural network model is trained according to the updated historical feature extraction database until the training of the convolutional neural network model meets the standards.

[0019] Furthermore, the real-time monitoring module divides the preset operation data group into a 70% operation training set, a 15% operation verification set and a 15% operation test set, inputs the operation training set into the decision tree model for training to obtain a trained decision tree model, inputs the operation verification set into the trained decision tree model to iteratively optimize the parameters of the trained decision tree model, inputs the operation test set into the iteratively optimized decision tree model, tests the iteratively optimized decision tree model to obtain an operation accuracy K, compares the operation accuracy K with the preset operation accuracy K0, judges the training qualification of the decision tree model according to the comparison result, and outputs the decision tree model according to the judgment result, wherein:

[0020] When K≥K0, the decision tree model is judged to be qualified and the decision tree model is output as a real-time monitoring model;

[0021] When K<K0, it is determined that the decision tree model training is unqualified, the preset operation data group is updated, and the decision tree model is repeatedly trained according to the updated preset operation data group until the decision tree model training meets the standard.

[0022] Furthermore, the real-time monitoring module also inputs the characteristic operation data into the real-time monitoring model for identification, obtains the identification result, judges the degradation state of the water storage unit according to the identification result, and outputs according to the judgment result, wherein:

[0023] When the identification result is that the operation is normal, the degradation state of the water storage unit is determined to be qualified, and the water storage unit is not maintained;

[0024] When the identification result is abnormal operation, the degradation state of the hydropower unit is determined to be unqualified, the characteristic operation data is compared with the preset maintenance template, and output is made according to the comparison result, where:

[0025] If there is preset characteristic operation data consistent with the characteristic operation data in the preset maintenance template, the preset maintenance plan corresponding to the preset characteristic data is output, and the water storage unit is maintained according to the preset maintenance plan;

[0026] If there is no preset characteristic operation data consistent with the characteristic operation data in the preset maintenance template, an alarm is issued to the administrator, who then performs maintenance on the hydropower unit and updates the preset maintenance template.

[0027] Furthermore, the degradation prediction module, when constructing the degradation prediction model, divides the preset prediction data group into a 70% prediction training set, a 20% prediction verification set and a 10% prediction test set, inputs the prediction training set into the LSTM model for training to obtain a trained LSTM model, inputs the prediction verification set into the trained LSTM model to iteratively optimize the parameters of the trained LSTM model, inputs the prediction test set into the iteratively optimized LSTM model, tests the iteratively optimized LSTM model to obtain a prediction accuracy L, compares the prediction accuracy L with the preset prediction accuracy L0, judges the training success of the LSTM model according to the comparison result, and outputs the LSTM model according to the judgment result, wherein:

[0028] When L≥L0, the LSTM model is judged to be successfully trained, and the LSTM model is output as a degradation prediction model;

[0029] When L<L0, it is determined that the LSTM model training is unsuccessful, the preset test data group is updated, and the LSTM model is repeatedly trained according to the updated preset test data group until the LSTM model training is successful.

[0030] Furthermore, when predicting the degradation state trend of the hydropower storage unit, the degradation prediction module inputs the characteristic operation data into the degradation prediction model for prediction, obtains the prediction result, judges the safety situation of the degradation state trend of the hydropower storage unit according to the prediction result, and outputs according to the judgment result, wherein:

[0031] When the prediction result shows that the trend is normal, the degradation state trend of the hydropower storage unit is determined to be safe, and no maintenance is performed on the hydropower storage unit;

[0032] When the prediction result is an abnormal trend, the degradation state trend of the hydropower unit is determined to be unsafe, and the characteristic operation data is compared with the preset maintenance template, and output is made according to the comparison result, where:

[0033] If there is preset characteristic operation data consistent with the characteristic operation data in the preset maintenance template, the preset maintenance plan corresponding to the preset characteristic data is output, and the water storage unit is maintained according to the preset maintenance plan;

[0034] If there is no preset characteristic operation data consistent with the characteristic operation data in the preset maintenance template, an alarm is issued to the administrator, who then performs maintenance on the hydropower unit and updates the preset maintenance template.

[0035] Furthermore, the prediction adjustment module obtains the prediction result of the degradation prediction module at each time node t and the recognition result of the real-time monitoring module at each time node t, compares the prediction result with the recognition result, judges the effectiveness of the degradation prediction module according to the comparison result, and adjusts the degradation prediction module according to the judgment result, wherein:

[0036] When the prediction result is consistent with the recognition result, the degradation prediction module is determined to be valid and no adjustment is made to the degradation prediction module;

[0037] When the prediction result is inconsistent with the identification result, the degradation prediction module is determined to be invalid, the degradation prediction module is adjusted, and a degradation prediction module failure alarm is issued to the administrator;

[0038] When adjusting the degradation prediction module, the prediction adjustment module resets the successfully trained LSTM model, updates the preset test data group, and repeatedly trains the LSTM model according to the updated preset test data group until the LSTM model is successfully trained.

[0039] Furthermore, the prediction optimization module obtains the number of adjustments H made by the prediction adjustment module to the degradation prediction module within the optimization period, compares the number of adjustments H with the preset number of adjustments H0, judges the practicality of the prediction adjustment module according to the comparison result, and optimizes the prediction adjustment module according to the judgment result, wherein:

[0040] When H<H0, it is determined that the practicality of the prediction adjustment module meets the standard, and the prediction adjustment module is not optimized;

[0041] When H≥H0, it is determined that the practicality of the prediction adjustment module does not meet the standard, and the judgment result of the prediction adjustment module is optimized to the failure of the degradation prediction module. An alarm that the practicality of the prediction adjustment module does not meet the standard is pushed to the administrator, and the administrator maintains the system.

[0042] On the other hand, the present invention provides a method for a rapid early warning system for a degradation state trend of a water storage unit, comprising:

[0043] Step S1, real-time collection of operating data of the water storage unit;

[0044] Step S2, performing feature processing on the operation data of the water storage unit to obtain feature operation data;

[0045] Step S3, constructing a real-time monitoring model and performing real-time monitoring on the degradation state of the water storage unit;

[0046] Step S4, constructing a degradation prediction model and predicting the degradation state trend of the hydropower storage unit;

[0047] Step S5, adjusting step S4 according to the prediction results and recognition results of each time node;

[0048] Step S6, optimizing step S5 according to the number of adjustments within the optimization cycle and the preset adjustment.

[0049] Compared with the prior art, the beneficial effect of the present invention lies in that the system collects the operation data of the water storage unit in real time through the data acquisition module, ensures the timeliness and accuracy of the data, and provides a rich data basis for subsequent feature processing and degradation prediction. The system pre-processes the collected operation data of the water storage unit through the feature processing module, removes abnormal values ​​and invalid data, ensures data quality, and thus improves the efficiency of data analysis and model training. The system constructs a model for real-time monitoring of the degradation state of the water storage unit through the real-time monitoring module to ensure the accuracy and reliability of real-time monitoring, and monitors the operation state of the water storage unit in real time, timely discovers and warns of potential degradation problems, and avoids the occurrence of failures. The system constructs and optimizes the LSTM for predicting the degradation state trend of the water storage unit through the degradation prediction module. The model ensures the accuracy and reliability of the prediction, and makes trend predictions on the degradation state of the hydropower unit, identifies possible degradation problems in advance, and performs preventive maintenance. The system dynamically adjusts the degradation prediction module according to the prediction results and real-time monitoring results of each time node t through the prediction adjustment module to ensure the continuous optimization and accuracy of the prediction model, enhance the adaptive ability of the system, and can dynamically adjust the prediction model according to real-time monitoring data and historical prediction data to improve the accuracy of the prediction. The system optimizes the prediction adjustment module according to the number of adjustments within the optimization cycle and the preset adjustment target through the prediction optimization module, reduces unnecessary adjustments, and improves system efficiency. By optimizing the adjustment strategy, the overall performance of the degradation prediction model is improved to ensure the stability and efficiency of the system in long-term operation.

[0050] In particular, the processing unit pre-processes the operating data of the hydropower unit by using a data editing tool, effectively removes noise, corrects abnormal values ​​and fills missing values, ensures the accuracy and completeness of the data, thereby reducing the consumption of computing resources and improving processing efficiency.

[0051] In particular, the feature extraction unit automates the feature extraction process, reduces manual intervention and human errors, improves efficiency and accuracy, and automatically updates the historical feature extraction database and retrains the convolutional neural network model when the convolutional neural network model training does not meet the standards until the convolutional neural network model meets the expected standards, thereby ensuring the continuous improvement and optimization of the convolutional neural network model. When the convolutional neural network model training meets the standards, it is used as a feature extraction model for feature extraction of preprocessed data, which can efficiently obtain feature operation data and improve the quality of subsequent data processing and analysis.

[0052] In particular, the real-time detection module ensures the high accuracy and reliability of the decision tree model through scientific data segmentation, iterative optimization, accuracy verification and automated training mechanisms. The automated judgment and adjustment process reduces manual intervention, improves the system's work efficiency, reduces maintenance costs, and enables the system to maintain efficient, stable and reliable real-time monitoring capabilities in a dynamic environment.

[0053] In particular, the real-time monitoring module can accurately judge the operating status of the water storage unit by inputting the characteristic operation data into the real-time monitoring model for identification, ensure that the system can detect the degradation status of the unit in time, and automatically judge the degradation status of the water storage unit according to the identification result, so as to reduce the error and uncertainty of manual judgment and improve the accuracy and efficiency of judgment. When the identification result is normal operation, the degradation status of the water storage unit is judged to be qualified and no maintenance is performed. When the identification result is abnormal operation, the degradation status of the unit is judged to be unqualified and maintenance is performed in time. If the characteristic operation data is consistent with the data in the preset maintenance template, the corresponding preset maintenance plan is automatically output to quickly respond to and handle abnormal situations and improve maintenance efficiency. If there is no data consistent with the characteristic operation data in the preset maintenance template, the system will issue an alarm to the administrator so that the administrator can intervene in time to perform necessary maintenance operations to prevent potential problems from expanding.

[0054] In particular, the degradation prediction module inputs the prediction test set into the iteratively optimized LSTM model for testing, obtains the prediction accuracy L, and compares it with the preset prediction accuracy L0, to ensure the accuracy of the LSTM model in practical applications and avoid overfitting or underfitting problems that may occur during the training of the LSTM model. When the prediction accuracy L is less than the preset accuracy L0, the system will update the preset test data group and retrain the LSTM model until the LSTM model reaches the expected accuracy, to ensure the adaptability and accuracy of the LSTM model, input the characteristic operation data into the degradation prediction model, and perform degradation state trend prediction. According to the prediction results, it is determined whether the degradation state trend of the hydropower unit is safe, so as to determine whether maintenance is required. When the prediction result is an abnormal trend, the system will further compare the characteristic operation data with the preset maintenance template. If there is consistent preset characteristic data, the corresponding preset maintenance plan will be output and executed to ensure rapid response and handling of abnormal situations. If there is no data consistent with the characteristic operation data in the preset maintenance template, the system will issue an alarm to the administrator, and the administrator can intervene in time for maintenance and update the maintenance template, thereby improving the flexibility and response speed of the system.

[0055] In particular, the prediction adjustment module can obtain the prediction results of the degradation prediction module and the identification results of the real-time monitoring module in real time, and compare them to ensure that the system can quickly identify the effectiveness of the degradation prediction module. When the prediction result is inconsistent with the identification result, the system will automatically adjust the degradation prediction module to improve the system's adaptability. When the degradation prediction module fails, the system will reset the successfully trained LSTM model and update the preset test data group, and then retrain the LSTM model until the training is successful to maintain high prediction accuracy. At the same time, when the degradation prediction module fails, the system will promptly send a failure alarm message to the administrator, which will help the administrator to take quick measures and reduce the risks that may be caused by system failure.

[0056] In particular, the prediction optimization module can automatically evaluate the practicality of the prediction adjustment module by comparing the adjustment times H and the preset adjustment times H0, which helps to maintain the efficiency and accuracy of the prediction model. When the practicality of the prediction adjustment module does not meet the standard, the judgment result of the prediction adjustment module is automatically optimized to the failure of the degraded prediction module, and an alarm information is pushed to the administrator to ensure that the system can take prompt measures when a problem is detected to reduce the risk of system failure. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 This is a structural schematic diagram of the rapid early warning system for the degradation trend of a water storage unit in this embodiment;

[0058] Figure 2 This is a schematic diagram of the structure of the feature processing module of this embodiment;

[0059] Figure 3 Schematic diagram of the process of the method for rapid early warning of degradation trend of a water storage unit in this embodiment. DETAILED DESCRIPTION

[0060] In order to make the objects and advantages of the present invention more clearly understood, the present invention is further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0061] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the protection scope of the present invention.

[0062] It should be noted that in the description of the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected" and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, or it can be the internal communication of two components. For those skilled in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0063] See also Figure 1 As shown, it is a structural schematic diagram of the rapid early warning system for the degradation state trend of the water storage unit in this example, and the system includes:

[0064] Data acquisition module, used to collect the operation data of the water storage unit in real time;

[0065] A feature processing module is used to perform feature processing on the operation data of the water storage unit to obtain feature operation data, and the feature processing module is connected to the data acquisition module;

[0066] A real-time monitoring module is used to construct a real-time monitoring model and to monitor the degradation state of the water storage unit in real time. The real-time monitoring module is connected to the feature processing module;

[0067] A degradation prediction module is used to construct a degradation prediction model and predict the degradation state trend of the hydropower storage unit. The degradation prediction module is connected to the feature processing module;

[0068] A prediction adjustment module, used to adjust the degradation prediction module according to the prediction results and identification results of each time node, and the prediction adjustment module is connected with the degradation prediction module and the real-time monitoring module;

[0069] The prediction optimization module is used to optimize the prediction adjustment module according to the number of adjustments within the optimization cycle and the preset adjustments. The prediction optimization module is connected to the prediction adjustment module.

[0070] Specifically, the system is applied to pumped storage power stations. The system can effectively improve the reliability and maintenance efficiency of the unit operation and reduce the failure rate and maintenance costs through real-time data collection, feature processing, real-time monitoring, degradation prediction and dynamic adjustment of the operation data of the water storage unit. In particular, the system collects the operation data of the water storage unit in real time through the data collection module to ensure the timeliness and accuracy of the data, and provide a rich data basis for subsequent feature processing and degradation prediction. The system pre-processes the collected operation data of the water storage unit through the feature processing module to remove outliers and invalid data to ensure data quality, thereby improving the efficiency of data analysis and model training. The system constructs a model for real-time monitoring of the degradation state of the water storage unit through the real-time monitoring module to ensure the accuracy and reliability of real-time monitoring, and monitors the operation state of the water storage unit in real time to promptly discover and warn of potential degradation problems and avoid failures. The occurrence of degradation, the system constructs and optimizes the LSTM model for predicting the degradation state trend of the water storage unit through the degradation prediction module, ensures the accuracy and reliability of the prediction, and predicts the trend of the degradation state of the water storage unit, identifies possible degradation problems in advance, and performs preventive maintenance. The system dynamically adjusts the degradation prediction module according to the prediction results and real-time monitoring results of each time node t through the prediction adjustment module, ensures the continuous optimization and accuracy of the prediction model, enhances the adaptive ability of the system, and can dynamically adjust the prediction model according to real-time monitoring data and historical prediction data to improve the accuracy of the prediction. The system optimizes the prediction adjustment module according to the number of adjustments within the optimization cycle and the preset adjustment target through the prediction optimization module, reduces unnecessary adjustments, and improves system efficiency. By optimizing the adjustment strategy, the overall performance of the degradation prediction model is improved to ensure the stability and efficiency of the system in long-term operation.

[0071] Specifically, the data acquisition module in this embodiment does not limit the real-time acquisition method of the operating data of the water storage unit. Relevant technical personnel in this field can freely set it according to actual needs, and only need to meet the needs of real-time acquisition of the operating data of the water storage unit. For example, the real-time acquisition method of the operating data of the water storage unit can be set to real-time acquisition of the operating data of the water storage unit through preset sensors. The preset sensors refer to preset sensor devices for real-time acquisition of the operating data of the water storage unit, such as current sensors, monitoring parameters, voltage sensors, temperature sensors, vibration sensors, pressure sensors, flow sensors, speed sensors and displacement sensors. The operating data of the water storage unit refers to parameters used to evaluate the operating status and performance of the unit and identify potential faults and degradation conditions, such as current data, voltage data, temperature data, vibration data, pressure data, flow data, speed data and displacement data.

[0072] See also Figure 2 As shown, it is a schematic diagram of the structure of the feature processing module of this embodiment, and the feature processing module includes:

[0073] A preprocessing unit, used to preprocess the operation data of the water storage unit to obtain preprocessed operation data;

[0074] The feature extraction unit is used to construct a feature extraction model and perform feature extraction on the preprocessing operation data according to the feature extraction model to obtain feature operation data. The feature extraction unit is connected to the preprocessing unit.

[0075] Specifically, the preprocessing unit preprocesses the operation data of the water storage unit through a data editing tool to obtain preprocessed operation data.

[0076] Specifically, the data editing tool refers to a tool used to pre-process the operation data of the water storage unit, such as an open source data sorting tool and Apache Spark, and the pre-processing refers to the processing process of the operation data of the water storage unit before inputting the operation data of the water storage unit into the feature extraction model, such as data cleaning, data conversion and missing value processing.

[0077] Specifically, the processing unit pre-processes the operating data of the hydropower unit by using a data editing tool, effectively removes noise, corrects abnormal values ​​and fills missing values, ensures the accuracy and completeness of the data, thereby reducing the consumption of computing resources and improving processing efficiency.

[0078] Specifically, when constructing a feature extraction model, the feature extraction unit divides the historical feature extraction database into a feature training set of 75%, a feature verification set of 15%, and a feature test set of 10%, inputs the feature training set into the convolutional neural network model to train the convolutional neural network model, inputs the feature verification set into the trained convolutional neural network model to iteratively optimize the parameters of the trained convolutional neural network model, and inputs the feature test set into the iteratively optimized convolutional neural network model for testing, sets the total number of samples in the feature test set to M0, the number of samples correctly output by the iteratively optimized convolutional neural network model to M1, and the feature accuracy to Q, Q=M1 / M0, compares the feature accuracy Q with the preset feature accuracy Q0, judges the training compliance of the convolutional neural network model according to the comparison result, and outputs according to the judgment result, wherein:

[0079] When Q≥Q0, it is determined that the convolutional neural network model training has reached the standard, the convolutional neural network model is output as a feature extraction model, and the preprocessed operation data is input into the feature extraction model for feature extraction to obtain feature operation data;

[0080] When Q<Q0, it is determined that the training of the convolutional neural network model does not meet the standards, the historical feature extraction database is updated, and the convolutional neural network model is initialized. The initialized convolutional neural network model is trained according to the updated historical feature extraction database until the training of the convolutional neural network model meets the standards.

[0081] Specifically, the historical feature extraction database refers to a database containing original data and original data after feature extraction for training the convolutional neural network model, the feature training set refers to a data set used to train the convolutional neural network model, the feature verification set refers to a data set used to adjust and optimize the parameters of the convolutional neural network model during the training process, the feature test set refers to a data set used to evaluate the training compliance of the convolutional neural network model after the convolutional neural network model training and verification are completed, the convolutional neural network model refers to a deep learning model used for feature extraction, the total number of samples in the feature test set refers to the total number of samples in the feature test set, the number of correctly output samples refers to the number of samples correctly classified and predicted by the convolutional neural network model in the feature test set, and the feature accuracy refers to the convolutional neural network The ratio of the number of samples correctly output by the model on the feature test set to the total number of samples in the feature test set. The preset feature accuracy refers to a pre-set value used to reflect the training compliance of the convolutional neural network model. This embodiment does not limit the specific value of the preset feature accuracy. Relevant technical personnel in this field can freely set it according to actual needs. It only needs to meet the need to judge the training compliance of the convolutional neural network model. For example, the preset feature accuracy Q0 can be set to 0.98. This embodiment does not limit the updating method of the historical feature extraction database. Relevant technical personnel in this field can freely set it according to actual needs. It only needs to meet the need to update the historical feature extraction database. For example, the updating method of the historical feature extraction database can be set to update the historical feature extraction database through big data.

[0082] Specifically, the feature extraction unit automates the feature extraction process, reduces manual intervention and human errors, improves efficiency and accuracy, and automatically updates the historical feature extraction database and retrains the convolutional neural network model when the convolutional neural network model training does not meet the standards until the convolutional neural network model meets the expected standards, thereby ensuring the continuous improvement and optimization of the convolutional neural network model. When the convolutional neural network model training meets the standards, it is used as a feature extraction model for feature extraction of preprocessed data, which can efficiently obtain feature operation data and improve the quality of subsequent data processing and analysis.

[0083] Specifically, the real-time monitoring module divides the preset operation data group into 70% of the operation training set, 15% of the operation verification set and 15% of the operation test set, inputs the operation training set into the decision tree model for training to obtain the trained decision tree model, inputs the operation verification set into the trained decision tree model to iteratively optimize the parameters of the trained decision tree model, and inputs the operation test set into the iteratively optimized decision tree model to test the iteratively optimized decision tree model to obtain the operation accuracy K, compares the operation accuracy K with the preset operation accuracy K0, judges the training qualification of the decision tree model according to the comparison result, and outputs the decision tree model according to the judgment result, wherein:

[0084] When K≥K0, the decision tree model is judged to be qualified and the decision tree model is output as a real-time monitoring model;

[0085] When K<K0, it is determined that the decision tree model training is unqualified, the preset operation data group is updated, and the decision tree model is repeatedly trained according to the updated preset operation data group until the decision tree model training meets the standard.

[0086] Specifically, the preset operation data group refers to a preset data group for training a decision tree model stored in the form of preset feature operation data-preset real-time monitoring results, preset feature operation data-preset real-time monitoring results, the decision tree model refers to a machine learning model for classification and regression tasks, the operation accuracy K refers to the proportion of the number of samples W correctly output by the convolutional neural network model to the total number of samples W0 in the operation test set, that is, K=W / W0, the preset operation accuracy K0 refers to a preset value for judging the qualified training of the decision tree model, this embodiment does not limit the value of the preset operation accuracy K0, and relevant technical personnel in this field can freely set it according to actual needs, and only need to meet the needs of judging the qualified training of the decision tree model, such as setting the preset operation accuracy K0=0.97.

[0087] Specifically, the real-time detection module ensures the high accuracy and reliability of the decision tree model through scientific data segmentation, iterative optimization, accuracy verification and automated training mechanisms. The automated judgment and adjustment processes reduce manual intervention, improve the system's work efficiency, and reduce maintenance costs, enabling the system to maintain efficient, stable and reliable real-time monitoring capabilities in a dynamic environment.

[0088] Specifically, the real-time monitoring module also inputs the characteristic operation data into the real-time monitoring model for identification, obtains the identification result, judges the degradation state of the water storage unit according to the identification result, and outputs according to the judgment result, wherein:

[0089] When the identification result is that the operation is normal, the degradation state of the water storage unit is determined to be qualified, and the water storage unit is not maintained;

[0090] When the identification result is abnormal operation, the degradation state of the hydropower unit is determined to be unqualified, the characteristic operation data is compared with the preset maintenance template, and output is made according to the comparison result, where:

[0091] If there is preset characteristic operation data consistent with the characteristic operation data in the preset maintenance template, the preset maintenance plan corresponding to the preset characteristic data is output, and the water storage unit is maintained according to the preset maintenance plan;

[0092] If there is no preset characteristic operation data consistent with the characteristic operation data in the preset maintenance template, an alarm is issued to the administrator, who then performs maintenance on the hydropower unit and updates the preset maintenance template.

[0093] Specifically, the preset maintenance template refers to a data set containing preset characteristic operation data and its corresponding preset maintenance plan, the preset characteristic operation data refers to pre-set characteristic operation data, the preset maintenance plan refers to a preset plan for maintaining the water storage unit, such as component replacement, lubrication management and electrical system maintenance, the administrator refers to a person who has maintenance responsibilities for the water storage unit and the system, and the alarm refers to an alarm signal that the system automatically generates and sends to the administrator when the system detects that there is no preset characteristic operation data consistent with the characteristic operation data in the preset maintenance template, such as pushing an alarm window to the administrator.

[0094] Specifically, the real-time monitoring module can accurately judge the operating status of the water storage unit by inputting the characteristic operation data into the real-time monitoring model for identification, ensure that the system can detect the degradation status of the unit in time, and automatically judge the degradation status of the water storage unit according to the identification result, so as to reduce the error and uncertainty of manual judgment and improve the accuracy and efficiency of judgment. When the identification result is normal operation, the degradation status of the water storage unit is judged to be qualified and no maintenance is performed. When the identification result is abnormal operation, the degradation status of the unit is judged to be unqualified and maintenance is performed in time. If the characteristic operation data is consistent with the data in the preset maintenance template, the corresponding preset maintenance plan is automatically output to quickly respond to and handle abnormal situations and improve maintenance efficiency. If there is no data consistent with the characteristic operation data in the preset maintenance template, the system will issue an alarm to the administrator so that the administrator can intervene in time to perform necessary maintenance operations to prevent potential problems from expanding.

[0095] Specifically, the degradation prediction module, when constructing the degradation prediction model, divides the preset prediction data group into a 70% prediction training set, a 20% prediction verification set and a 10% prediction test set, inputs the prediction training set into the LSTM model for training to obtain the trained LSTM model, inputs the prediction verification set into the trained LSTM model to iteratively optimize the parameters of the trained LSTM model, inputs the prediction test set into the iteratively optimized LSTM model, tests the iteratively optimized LSTM model to obtain the prediction accuracy L, compares the prediction accuracy L with the preset prediction accuracy L0, judges the training success of the LSTM model according to the comparison result, and outputs the LSTM model according to the judgment result, wherein:

[0096] When L≥L0, the LSTM model is judged to be successfully trained, and the LSTM model is output as a degradation prediction model;

[0097] When L<L0, it is determined that the LSTM model training is unsuccessful, the preset test data group is updated, and the LSTM model is repeatedly trained according to the updated preset test data group until the LSTM model training is successful;

[0098] The degradation prediction module, when predicting the degradation state trend of the hydropower storage unit, inputs the characteristic operation data into the degradation prediction model for prediction, obtains the prediction result, judges the safety situation of the degradation state trend of the hydropower storage unit according to the prediction result, and outputs according to the judgment result, wherein:

[0099] When the prediction result shows that the trend is normal, the degradation state trend of the hydropower storage unit is determined to be safe, and no maintenance is performed on the hydropower storage unit;

[0100] When the prediction result is an abnormal trend, the degradation state trend of the hydropower unit is determined to be unsafe, and the characteristic operation data is compared with the preset maintenance template, and output is made according to the comparison result, where:

[0101] If there is preset characteristic operation data consistent with the characteristic operation data in the preset maintenance template, the preset maintenance plan corresponding to the preset characteristic data is output, and the water storage unit is maintained according to the preset maintenance plan;

[0102] If there is no preset characteristic operation data consistent with the characteristic operation data in the preset maintenance template, an alarm is issued to the administrator, who then performs maintenance on the hydropower unit and updates the preset maintenance template.

[0103] Specifically, the LSTM model refers to a recursive neural network used to process and predict long-term dependencies in time series data, the preset prediction data group refers to a preset database for training the LSTM model, the prediction training set refers to a data set used for training the LSTM model, the prediction verification set refers to a data set used to optimize the parameters of the LSTM model during the training process, and the prediction test set refers to a data set used to evaluate the performance of the LSTM model.

[0104] Specifically, the degradation prediction module inputs the prediction test set into the iteratively optimized LSTM model for testing, obtains the prediction accuracy L, and compares it with the preset prediction accuracy L0 to ensure the accuracy of the LSTM model in practical applications and avoid overfitting or underfitting problems that may occur during the training of the LSTM model. When the prediction accuracy L is less than the preset accuracy L0, the system will update the preset test data group and retrain the LSTM model until the LSTM model reaches the expected accuracy to ensure the adaptability and accuracy of the LSTM model. The characteristic operation data is input into the degradation prediction model to predict the degradation state trend. According to the prediction results, it is judged whether the degradation state trend of the water storage unit is safe, so as to decide whether maintenance is needed. When the prediction result is an abnormal trend, the system will further compare the characteristic operation data with the preset maintenance template. If there is consistent preset characteristic data, the corresponding preset maintenance plan will be output and executed to ensure rapid response and handling of abnormal situations. If there is no data consistent with the characteristic operation data in the preset maintenance template, the system will issue an alarm to the administrator, and the administrator can intervene in time for maintenance and update the maintenance template, thereby improving the flexibility and response speed of the system.

[0105] Specifically, the prediction adjustment module obtains the prediction result of the degradation prediction module for each time node t and the recognition result of the real-time monitoring module at each time node t, compares the prediction result with the recognition result, judges the effectiveness of the degradation prediction module according to the comparison result, and adjusts the degradation prediction module according to the judgment result, wherein:

[0106] When the prediction result is consistent with the recognition result, the degradation prediction module is determined to be valid and no adjustment is made to the degradation prediction module;

[0107] When the prediction result is inconsistent with the identification result, the degradation prediction module is determined to be invalid, the degradation prediction module is adjusted, and a degradation prediction module failure alarm is issued to the administrator;

[0108] When adjusting the degradation prediction module, the prediction adjustment module resets the successfully trained LSTM model, updates the preset test data group, and repeatedly trains the LSTM model according to the updated preset test data group until the LSTM model is successfully trained.

[0109] Specifically, each time node t refers to a time point at which the system performs prediction and monitoring, and the consistency between the prediction result and the identification result means that the prediction result of the degradation prediction module and the identification result of the real-time monitoring module are consistent at each time node t, for example: if the degradation prediction module predicts that the trend is normal at time point t1, the identification result of the real-time monitoring module at time point t1 is normal operation; if the degradation prediction module predicts that the trend is abnormal at time point t1, the identification result of the real-time monitoring module at time point t1 is abnormal operation, and the inconsistency between the prediction result and the identification result means that the prediction result of the degradation prediction module and the identification result of the real-time monitoring module are inconsistent at each time node t, for example: if the degradation prediction module predicts that the trend is abnormal at time point t2, but the identification result of the real-time monitoring module at time point t2 is normal operation; if the degradation prediction module predicts that the trend is normal at time point t2, but the identification result of the real-time monitoring module at time point t2 is abnormal operation.

[0110] Specifically, the prediction adjustment module can obtain the prediction results of the degradation prediction module and the identification results of the real-time monitoring module in real time, and compare them to ensure that the system can quickly identify the effectiveness of the degradation prediction module. When the prediction result is inconsistent with the identification result, the system will automatically adjust the degradation prediction module to improve the system's adaptability. When the degradation prediction module fails, the system will reset the successfully trained LSTM model and update the preset test data group, and then retrain the LSTM model until the training is successful to maintain high prediction accuracy. At the same time, when the degradation prediction module fails, the system will promptly send a failure alarm message to the administrator, which will help the administrator to take quick measures and reduce the risks that may be caused by system failure.

[0111] Specifically, the prediction optimization module obtains the number of adjustments H made by the prediction adjustment module to the degradation prediction module within the optimization period, compares the number of adjustments H with the preset number of adjustments H0, judges the practicality of the prediction adjustment module according to the comparison result, and optimizes the prediction adjustment module according to the judgment result, wherein:

[0112] When H<H0, it is determined that the practicality of the prediction adjustment module meets the standard, and the prediction adjustment module is not optimized;

[0113] When H≥H0, it is determined that the practicality of the prediction adjustment module does not meet the standard, and the judgment result of the prediction adjustment module is optimized to the failure of the degradation prediction module. An alarm that the practicality of the prediction adjustment module does not meet the standard is pushed to the administrator, and the administrator maintains the system.

[0114] Specifically, the optimization cycle refers to the preset time interval for evaluating the prediction adjustment module. For example, if the optimization cycle is 7 days, the number of adjustments H made by the prediction adjustment module to the degradation prediction module within 7 days is obtained. The number of adjustments H refers to the total number of times the prediction adjustment module adjusts the degradation prediction module within the optimization cycle. The preset number of adjustments H0 refers to a preset value for judging the practicality of the prediction adjustment module. This embodiment does not limit the specific value of the preset number of adjustments H0. Relevant technical personnel in this field can freely set it according to actual needs, and only need to meet the need to judge whether the practicality of the prediction adjustment module meets the standards. For example, the preset number of adjustments H0 can be set to 3.

[0115] Specifically, the prediction optimization module can automatically evaluate the practicality of the prediction adjustment module by comparing the adjustment times H and the preset adjustment times H0, which helps to maintain the efficiency and accuracy of the prediction model. When the practicality of the prediction adjustment module does not meet the standard, the judgment result of the prediction adjustment module is automatically optimized to the failure of the degraded prediction module, and an alarm information is pushed to the administrator to ensure that the system can take prompt measures when a problem is detected to reduce the risk of system failure.

[0116] See also Figure 3 As shown, it is a flow chart of a method for rapid early warning of degradation trend of a water storage unit in this embodiment, and the method comprises:

[0117] Step S1, real-time collection of operating data of the water storage unit;

[0118] Step S2, performing feature processing on the operation data of the water storage unit to obtain feature operation data;

[0119] Step S3, constructing a real-time monitoring model and performing real-time monitoring on the degradation state of the water storage unit;

[0120] Step S4, constructing a degradation prediction model and predicting the degradation state trend of the hydropower storage unit;

[0121] Step S5, adjusting step S4 according to the prediction results and recognition results of each time node;

[0122] Step S6, optimizing step S5 according to the number of adjustments within the optimization cycle and the preset adjustment.

[0123] So far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.

Claims

1. A rapid early warning system for degradation trend of a hydropower storage unit, characterized in that: include: Data acquisition module, used to collect the operation data of the water storage unit in real time; A feature processing module is used to perform feature processing on the operation data of the water storage unit to obtain feature operation data; Real-time monitoring module, used to build a real-time monitoring model and monitor the degradation status of the water storage unit in real time; Deterioration prediction module, used to build a degradation prediction model and predict the degradation state trend of the hydropower storage unit; A prediction adjustment module, used to adjust the degradation prediction module according to the prediction results and identification results of each time node; The prediction optimization module is used to optimize the prediction adjustment module according to the number of adjustments within the optimization cycle and the preset adjustments.

2. The rapid early warning system for degradation trend of a water storage unit according to claim 1 is characterized in that: The feature processing module comprises: A preprocessing unit, used to preprocess the operation data of the water storage unit to obtain preprocessed operation data; The feature extraction unit is used to construct a feature extraction model and perform feature extraction on the preprocessing operation data according to the feature extraction model to obtain feature operation data. The feature extraction unit is connected to the preprocessing unit.

3. The rapid early warning system for degradation trend of a water storage unit according to claim 2 is characterized in that: The preprocessing unit preprocesses the operation data of the water storage unit through a data editing tool to obtain preprocessed operation data; The feature extraction unit, when constructing a feature extraction model, divides the historical feature extraction database into a feature training set of 75%, a feature verification set of 15%, and a feature test set of 10%, inputs the feature training set into the convolutional neural network model to train the convolutional neural network model, inputs the feature verification set into the trained convolutional neural network model to iteratively optimize the parameters of the trained convolutional neural network model, and inputs the feature test set into the iteratively optimized convolutional neural network model for testing, sets the total number of samples in the feature test set to M0, the number of samples correctly output by the iteratively optimized convolutional neural network model to M1, and the feature accuracy to Q, Q=M1 / M0, compares the feature accuracy Q with the preset feature accuracy Q0, judges the training compliance of the convolutional neural network model according to the comparison result, and outputs according to the judgment result, wherein: When Q≥Q0, it is determined that the convolutional neural network model training has reached the standard, the convolutional neural network model is output as a feature extraction model, and the preprocessed operation data is input into the feature extraction model for feature extraction to obtain feature operation data; When Q<Q0, it is determined that the training of the convolutional neural network model does not meet the standards, the historical feature extraction database is updated, and the convolutional neural network model is initialized. The initialized convolutional neural network model is trained according to the updated historical feature extraction database until the training of the convolutional neural network model meets the standards.

4. The rapid early warning system for degradation trend of a water storage unit according to claim 1 is characterized in that: The real-time monitoring module divides the preset operation data group into 70% of the operation training set, 15% of the operation verification set and 15% of the operation test set, inputs the operation training set into the decision tree model for training to obtain the trained decision tree model, inputs the operation verification set into the trained decision tree model to iteratively optimize the parameters of the trained decision tree model, and inputs the operation test set into the iteratively optimized decision tree model to test the iteratively optimized decision tree model to obtain the operation accuracy K, compares the operation accuracy K with the preset operation accuracy K0, judges the training qualification of the decision tree model according to the comparison result, and outputs the decision tree model according to the judgment result, wherein: When K≥K0, the decision tree model is judged to be qualified and the decision tree model is output as a real-time monitoring model; When K<K0, it is determined that the decision tree model training is unqualified, the preset operation data group is updated, and the decision tree model is repeatedly trained according to the updated preset operation data group until the decision tree model training meets the standard.

5. The rapid early warning system for degradation trend of a water storage unit according to claim 1 is characterized in that: The real-time monitoring module also inputs the characteristic operation data into the real-time monitoring model for identification, obtains the identification result, judges the degradation state of the water storage unit according to the identification result, and outputs according to the judgment result, wherein: When the identification result is that the operation is normal, the degradation state of the water storage unit is determined to be qualified, and no maintenance is performed on the water storage unit; When the identification result is abnormal operation, the degradation state of the hydropower unit is determined to be unqualified, the characteristic operation data is compared with the preset maintenance template, and output is made according to the comparison result, where: If there is preset characteristic operation data consistent with the characteristic operation data in the preset maintenance template, the preset maintenance plan corresponding to the preset characteristic data is output, and the water storage unit is maintained according to the preset maintenance plan; If there is no preset characteristic operation data consistent with the characteristic operation data in the preset maintenance template, an alarm is issued to the administrator, who then performs maintenance on the hydropower unit and updates the preset maintenance template.

6. The rapid early warning system for degradation trend of a water storage unit according to claim 1 is characterized in that: The degradation prediction module, when constructing the degradation prediction model, divides the preset prediction data group into a 70% prediction training set, a 20% prediction verification set and a 10% prediction test set, inputs the prediction training set into the LSTM model for training to obtain a trained LSTM model, inputs the prediction verification set into the trained LSTM model to iteratively optimize the parameters of the trained LSTM model, inputs the prediction test set into the iteratively optimized LSTM model, tests the iteratively optimized LSTM model to obtain a prediction accuracy L, compares the prediction accuracy L with a preset prediction accuracy L0, judges the training success of the LSTM model according to the comparison result, and outputs the LSTM model according to the judgment result, wherein: When L≥L0, the LSTM model is judged to be successfully trained, and the LSTM model is output as a degradation prediction model; When L<L0, it is determined that the LSTM model training is unsuccessful, the preset test data group is updated, and the LSTM model is repeatedly trained according to the updated preset test data group until the LSTM model training is successful.

7. The rapid early warning system for degradation trend of a water storage unit according to claim 1 is characterized in that: The degradation prediction module, when predicting the degradation state trend of the hydropower storage unit, inputs the characteristic operation data into the degradation prediction model for prediction, obtains the prediction result, judges the safety situation of the degradation state trend of the hydropower storage unit according to the prediction result, and outputs according to the judgment result, wherein: When the prediction result shows that the trend is normal, the degradation state trend of the hydropower storage unit is determined to be safe, and no maintenance is performed on the hydropower storage unit; When the prediction result is an abnormal trend, the degradation state trend of the hydropower unit is determined to be unsafe, and the characteristic operation data is compared with the preset maintenance template, and output is made according to the comparison result, where: If there is preset characteristic operation data consistent with the characteristic operation data in the preset maintenance template, the preset maintenance plan corresponding to the preset characteristic data is output, and the water storage unit is maintained according to the preset maintenance plan; If there is no preset characteristic operation data consistent with the characteristic operation data in the preset maintenance template, an alarm is issued to the administrator, who then performs maintenance on the hydropower unit and updates the preset maintenance template.

8. The rapid early warning system for degradation trend of a water storage unit according to claim 1 is characterized in that: The prediction adjustment module obtains the prediction result of the degradation prediction module at each time node t and the recognition result of the real-time monitoring module at each time node t, compares the prediction result with the recognition result, judges the effectiveness of the degradation prediction module according to the comparison result, and adjusts the degradation prediction module according to the judgment result, wherein: When the prediction result is consistent with the recognition result, the degradation prediction module is determined to be valid and no adjustment is made to the degradation prediction module; When the prediction result is inconsistent with the identification result, the degradation prediction module is determined to be invalid, the degradation prediction module is adjusted, and a degradation prediction module failure alarm is issued to the administrator; When adjusting the degradation prediction module, the prediction adjustment module resets the successfully trained LSTM model, updates the preset test data group, and repeatedly trains the LSTM model according to the updated preset test data group until the LSTM model is successfully trained.

9. The rapid early warning system for degradation trend of a water storage unit according to claim 1 is characterized in that: The prediction optimization module obtains the number of adjustment times H of the prediction adjustment module for the degradation prediction module within the optimization period, compares the number of adjustment times H with the preset number of adjustment times H0, judges the practicality of the prediction adjustment module according to the comparison result, and optimizes the prediction adjustment module according to the judgment result, wherein: When H<H0, it is determined that the practicality of the prediction adjustment module meets the standard, and the prediction adjustment module is not optimized; When H≥H0, it is determined that the practicality of the prediction adjustment module does not meet the standard, and the judgment result of the prediction adjustment module is optimized to the failure of the degradation prediction module. An alarm that the practicality of the prediction adjustment module does not meet the standard is pushed to the administrator, and the administrator maintains the system.

10. A method for the rapid early warning system for degradation trend of a water storage unit as claimed in claims 1 to 9, characterized in that: include: Step S1, real-time collection of operating data of the water storage unit; Step S2, performing feature processing on the operation data of the water storage unit to obtain feature operation data; Step S3, constructing a real-time monitoring model and performing real-time monitoring on the degradation state of the water storage unit; Step S4, constructing a degradation prediction model and predicting the degradation state trend of the hydropower storage unit; Step S5, adjusting step S4 according to the prediction results and recognition results of each time node; Step S6, optimizing step S5 according to the number of adjustments within the optimization cycle and the preset adjustment.

Images