A method and system for constructing an intelligent early warning model for hydropower data

Through the construction method of intelligent early warning model for hydropower data, the problem of low efficiency and low accuracy of early warning methods for hydropower station data is solved. Through data preprocessing and time series data prediction network architecture, model parameters are optimized, and more accurate and timely early warning is achieved, which improves the Operational safety and operational efficiency of hydropower stations.

CN118535862BActive Publication Date: 2024-11-22SANXIA JINSHAJIANG YUNCHUAN HYDROPOWER DEV CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411003014.9
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-07-25
Publication Date
2024-11-22
Estimated Expiration
2044-07-25

AI Technical Summary

Technical Problem

The existing hydropower station data warning methods are low efficiency, low accuracy, poor generalization ability, and difficult to effectively predict abnormal operation of hydropower units and provide accurate warnings. Especially in the context of big data, how to extract valuable information from massive data for accurate timing. There are challenges in forecasting and early warning.

Method used

A method of intelligent early warning model for hydropower data is adopted, including collecting operation data of hydropower units, screening steady-state operation data and stable operation areas, building a data preprocessing mechanism, setting epoch and batchsize based on the network architecture of time series data prediction, and building the best foundation and the model is tested and fine-tuned and performance evaluation is performed to optimize the intelligent early warning model of a single unit.

Benefits of technology

By eliminating noise and unstable data points, improve data accuracy and representativeness, improve model prediction accuracy and reliability, dynamically adjust model parameters to adapt to data sets of different sizes, and improve training flexibility and generalization capabilities. , ensure the accuracy and timeliness of early warnings, reduce false alarm rates, and improve response speed and accuracy.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118535862B_ABST
    Figure CN118535862B_ABST
Patent Text Reader

Abstract

The present invention discloses a method and system for constructing an intelligent early warning model for hydropower data, which relates to the technical field of intelligent hydropower station management, including collecting the operating data of hydropower units, screening steady-state operating data and stable operating areas, and constructing a unit data preprocessing mechanism; setting epoch and batchsize based on a network architecture for time series data prediction, and constructing an optimal base model; testing and fine-tuning the optimal base model, obtaining a single unit intelligent early warning model, and evaluating the performance of the hydropower unit and optimizing the single unit intelligent early warning model. The method of the present invention makes the early warning model more in line with the actual operating conditions and needs by performing detailed testing and fine-tuning on the model, thereby improving the accuracy and timeliness of the early warning, and realizing personalized adaptation of the model by applying the model to a single unit and optimizing it, thereby ultimately reducing the false alarm rate and improving the early warning response speed and accuracy.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent hydropower station management, and in particular to a method and system for constructing an intelligent early warning model for hydropower data. Background Art

[0002] Over the past few decades, the management and operation of hydropower stations have undergone a transformation from manual monitoring to automated systems and then to intelligent management. In the early days, hydropower stations relied on manual inspections and basic mechanical instruments to monitor operating status. With the development of computing technology and automatic control technology, automated monitoring systems have been widely deployed, which can collect and process the operating data of hydropower units in real time. In recent years, with the rise of big data, cloud computing and artificial intelligence technologies, intelligent hydropower station management systems have begun to emerge. These systems are capable of more complex data analysis and improve the operating efficiency and safety level of hydropower stations.

[0003] Although the existing automation and intelligent management system has improved the operating efficiency and safety of hydropower stations to a certain extent, there are still some shortcomings. For example, the existing system has high requirements for equipment data (quantity and quality) when building equipment models in batches, because if the data volume of the deep learning algorithm is not large enough, it is easy to overfit. When the existing construction method faces multiple devices, the construction time is very long and it is necessary to start training from random parameters. Our invention selects a suitable initial parameter. The existing system needs to be improved in processing complex time series data, predicting abnormal operation of hydropower units and making accurate early warnings. Especially in the context of big data, how to effectively extract valuable information from massive operating data, make accurate time series predictions, and implement early warnings based on this has become a problem that the existing technology urgently needs to solve. In addition, the existing technology also has limitations in data preprocessing, feature extraction and model optimization. These limitations may lead to slow response speed, low accuracy or poor generalization ability of the early warning system. Summary of the invention

[0004] In view of the above-mentioned problems, the present invention is proposed.

[0005] Therefore, the technical problem solved by the present invention is that the existing hydropower station data early warning method has low efficiency, low accuracy, poor generalization ability, and the problem of how to predict abnormal operation of hydropower units and make accurate early warnings.

[0006] To solve the above technical problems, the present invention provides the following technical solutions: a method for constructing an intelligent early warning model for hydropower data, comprising collecting operating data of hydropower units, screening steady-state operating data and stable operating areas, and constructing a unit data preprocessing mechanism; setting epoch and batchsize based on a network architecture for time series data prediction, and constructing an optimal base model; testing and fine-tuning the optimal base model to obtain an intelligent early warning model for a single unit, performing performance evaluation on the hydropower unit, and optimizing the intelligent early warning model for a single unit.

[0007] As a preferred scheme of the method for constructing an intelligent early warning model for hydropower data described in the present invention, the operating data of the hydropower unit includes operating condition data, bearing vibration data, hydrological data and power grid demand data; the operating condition data includes the speed, flow, pressure and temperature of the turbine; the bearing vibration data includes vibration frequency and amplitude.

[0008] As a preferred solution of the method for constructing the hydropower data intelligent early warning model of the present invention, the screening of steady-state operation data and stable operation area includes screening steady-state operation data and stable operation area based on active power, water head and fixed attributes of the unit; the steady-state operation data constraint is expressed as:

[0009] ;

[0010] in, Indicates at time Initially, the window size is The power fluctuation range within Indicates at time arrive The maximum value of internal power, Indicates at time arrive The minimum value of internal power, Indicates the window size Auxiliary variables for internal time traversal, screening any time point in a 10-minute window where the power fluctuation does not exceed 10MW The stable operation zone is a fixed interval. Within the stable operation zone, the various performance indicators of the unit meet the stable operation requirements of the power system. The fixed stable operation zone range of each unit is screened by active power and water head.

[0011] As a preferred solution of the method for constructing the hydropower data intelligent early warning model of the present invention, the unit data preprocessing mechanism includes data standardization, abnormal data screening and data splicing of the collected operating data of the hydropower unit; abnormal data screening includes marking the values ​​lasting more than 3 minutes as dead values, screening the dead values, and standardizing the operating data of the hydropower unit, which is expressed as:

[0012] ;

[0013] in, Represented as the i-th standardized data, is the i-th original data, and are the mean and standard deviation of the original data, respectively, and the weight of the hydropower unit operation characteristics is introduced , identify and filter outliers, expressed as:

[0014] ;

[0015] in, is the optimized local anomaly factor value, is any observation point in the data set, is the i-th data point of The nearest neighbor set, is the i-th data point The local reachable density of , reflecting the i-th data point The density of the surrounding area, is the local reachable density of any observation point, when When the i-th data point Mark outliers and filter them out; splice the operating data of the hydropower units into a comprehensive data set.

[0016] As a preferred solution of the method for constructing the hydropower data intelligent early warning model of the present invention, the construction of the optimal base model includes: The input data at the moment is used for feature extraction, and the feature extraction objective function It is expressed as:

[0017] ;

[0018] in, is the feature extraction weight, is the bias in feature extraction, for Input data at each moment, dynamically adjust epoch and batchsize, build a network architecture for time series data prediction based on feature extraction process, and the network architecture objective function It is expressed as:

[0019] ;

[0020] in, is the number of model training times, represents the size of a subset selected from the comprehensive dataset, for The adjustment coefficient, for The adjustment coefficient, is the length of the time series, The weight parameters that are dynamically adjusted for the network architecture, Dynamically adjust the bias parameters of the network architecture; Based on the network architecture predicted by the operation data and time series data of the hydropower unit, the nonlinear characteristics of the operating data are processed, the deviation degree of the bearing runout data is calculated, and the optimal base model is constructed , expressed as:

[0021] ;

[0022] in, is the normalization factor, is the adjustment coefficient affecting the network architecture objective function, is the total number of parameters in the operating condition data, The working condition data The weight of the parameters, Indicates The value of the operating parameter, is the total number of parameters in the bearing runout data, Indicates Observed values ​​of bearing runout data, Indicates The average value of the bearing runout data.

[0023] As a preferred solution of the method for constructing the hydropower data intelligent early warning model described in the present invention, the method of obtaining the intelligent early warning model of a single unit includes performing sample testing on the optimal base model and performing network architecture objective function In and Adjust and output the optimal single unit intelligent early warning model ; When the sample size is less than or equal to 10, Adjusted to 1 for online learning, the best base model learns features from each sample.

[0024] As a preferred solution of the method for constructing a hydropower data intelligent early warning model described in the present invention, wherein: the performance evaluation of the hydropower unit and optimization of the single unit intelligent early warning model includes evaluating the performance of the hydropower unit based on the output value M of the optimal single unit intelligent early warning model. When M is greater than 0 and less than or equal to 0.6, the hydropower station is in good operating condition and does not trigger the hydropower data early warning; when M is greater than 0.6 and less than or equal to 1, there is a problem with the operating condition of the hydropower station, which triggers the hydropower data early warning and sends out an early warning signal. The early warning system prompts the operation and maintenance personnel to check the unit status, identify and solve the problem; by balancing the training speed and memory usage efficiency of the single unit intelligent early warning model, Perform dynamic adjustment and optimization, output Optimal value.

[0025] Another object of the present invention is to provide a hydropower data intelligent early warning model construction system, which can set epoch and batchsize through a network architecture based on time series data prediction, build an optimal base model, and solve the current problem of low accuracy of hydropower station data early warning.

[0026] As a preferred solution of the hydropower data intelligent early warning model construction system described in the present invention, it includes: a data processing module, a base construction module, and a performance evaluation module; the data processing module is used to collect the operating data of the hydropower unit, screen the steady-state operating data and the stable operating area, and build a unit data preprocessing mechanism; the base construction module is used to set the epoch and batchsize based on the network architecture predicted by time series data, and build the optimal base model; the performance evaluation module is used to test and fine-tune the optimal base model, obtain a single unit intelligent early warning model, perform performance evaluation on the hydropower unit and optimize the single unit intelligent early warning model.

[0027] A computer device includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement a step of a method for constructing an intelligent early warning model for hydropower data.

[0028] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of a method for constructing an intelligent early warning model for hydropower data.

[0029] Beneficial effects of the present invention: The method for constructing an intelligent early warning model for hydropower data provided by the present invention ensures that the data entering the model has higher accuracy and representativeness by eliminating noise and unstable data points. In addition, data screening in the stable operation area helps to focus on the normal working state of the unit, thereby improving the accuracy and reliability of model prediction. By fine-tuning the parameters of model training, the learning process is optimized, ensuring that the model can effectively learn and capture complex patterns in time series data. Dynamic adjustment of epoch and batchsize not only enables the model to adapt to data sets of different sizes, but also improves the flexibility of training and the generalization ability of the model. By carefully testing and fine-tuning the model, it is ensured that the model achieves optimal performance in the application of a specific unit. The performance evaluation and optimization in this process make the early warning model more in line with the actual operating conditions and needs, and improve the accuracy and timeliness of the early warning. By applying the model to a single unit and optimizing it, the personalized adaptation of the model is realized, and customized early warning services can be provided according to the characteristics and operating conditions of the specific unit, ultimately reducing the false alarm rate and improving the early warning response speed and accuracy. The present invention achieves better results in terms of flexibility, reliability and response speed. BRIEF DESCRIPTION OF THE DRAWINGS

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

[0031] Figure 1 An overall flow chart of a method for constructing an intelligent early warning model for hydropower data provided in the first embodiment of the present invention.

[0032] Figure 2 An overall flow chart of a hydropower data intelligent early warning model construction system provided for the third embodiment of the present invention. DETAILED DESCRIPTION

[0033] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0034] Example 1

[0035] Reference Figure 1, which is an embodiment of the present invention, provides a method for constructing a hydropower data intelligent early warning model, comprising:

[0036] S1: Collect the operating data of the hydropower units, screen the steady-state operating data and stable operating area, and build a unit data preprocessing mechanism.

[0037] Furthermore, the operating data of the hydropower unit includes operating condition data, bearing vibration data, hydrological data and grid demand data; the operating condition data includes the speed, flow, pressure and temperature of the turbine; the bearing vibration data includes vibration frequency and amplitude.

[0038] It should be noted that screening steady-state operation data and stable operation area includes screening steady-state operation data and stable operation area based on active power, water head and fixed attributes of the unit; the steady-state operation data constraint is expressed as:

[0039] ;

[0040] in, Indicates at time Initially, the window size is The power fluctuation range within Indicates at time arrive The maximum value of internal power, Indicates at time arrive The minimum value of internal power, Indicates the window size Auxiliary variables for internal time traversal, screening any time point in a 10-minute window where the power fluctuation does not exceed 10MW The stable operation zone is a fixed interval. Within the stable operation zone, the various performance indicators of the unit meet the stable operation requirements of the power system. The fixed stable operation zone range of each unit is screened by active power and water head.

[0041] It should also be noted that the construction of the unit data preprocessing mechanism includes data standardization, abnormal data screening and data splicing of the collected hydropower unit operation data; abnormal data screening includes marking the values ​​lasting more than 3 minutes as dead values, screening out the dead values, and standardizing the hydropower unit operation data, which is expressed as:

[0042] ;

[0043] in, Represented as the i-th standardized data, is the i-th original data, and are the mean and standard deviation of the original data, respectively, and the weight of the hydropower unit operation characteristics is introduced , identify and filter outliers, expressed as:

[0044] ;

[0045] in, is the optimized local anomaly factor value, is any observation point in the data set, is the i-th data point of The nearest neighbor set, is the i-th data point The local reachable density of , reflecting the i-th data point The density of the surrounding area, is the local reachable density of any observation point, when When the i-th data point Mark outliers and filter them out; splice the operating data of the hydropower units into a comprehensive data set.

[0046] It should also be noted that the operating data of multiple units are collected at the same time, and a unified data preprocessing pipeline is built for units of different models. This can save a lot of data preprocessing time. Through the refined data screening process, the data of the unit in a stable operating state can be effectively identified, thereby providing a more accurate and stable data basis for subsequent model training. The constraints on steady-state operating data, such as the limitation of the power fluctuation range, not only reflect the in-depth understanding of the stable operation requirements of the power system, but also ensure the scientificity and rationality of data processing. The construction of a unit data preprocessing mechanism, including data standardization and abnormal data screening, not only improves the efficiency of data processing, but also introduces weights and the optimized local anomaly factor value , effectively identifying and eliminating outliers, improving data quality, taking into account the standardized processing of data, and adding an outlier screening mechanism based on the weight of operating characteristics, thereby improving the reliability and representativeness of the pre-processed data.

[0047] S2: Based on the network architecture for time series data prediction, set epoch and batchsize to build the best base model.

[0048] Furthermore, constructing the optimal base model involves concentrating The input data at the moment is used for feature extraction, and the feature extraction objective function It is expressed as:

[0049] ;

[0050] in, is the feature extraction weight, is the bias in feature extraction, for Input data at each moment, dynamically adjust epoch and batchsize, build a network architecture for time series data prediction based on feature extraction process, and the network architecture objective function It is expressed as:

[0051] ;

[0052] in, is the number of times the model is trained, represents the size of a subset selected from the comprehensive dataset, for The adjustment coefficient, for The adjustment coefficient, is the length of the time series, The weight parameters that are dynamically adjusted for the network architecture, Dynamically adjust the bias parameters of the network architecture; Based on the network architecture predicted by the operation data and time series data of the hydropower unit, the nonlinear characteristics of the operating data are processed, the deviation degree of the bearing runout data is calculated, and the optimal base model is constructed , expressed as:

[0053] ;

[0054] in, is the normalization factor, is the adjustment coefficient affecting the network architecture objective function, is the total number of parameters in the operating condition data, The working condition data The weight of the parameters, Indicates The value of the operating parameter, is the total number of parameters in the bearing runout data, Indicates Observed values ​​of bearing runout data, Indicates The average value of the bearing runout data.

[0055] It should be noted that feature extraction helps alleviate the gradient vanishing problem and improve the learning efficiency of the model. The network architecture objective function takes into account the number of training times. and subset size Dynamic adjustment of model performance, where and As an adjustment coefficient, it affects the adaptability of the model to different training strategies. It aims to find the learning strategy that best suits the data characteristics by dynamically adjusting the key parameters of the learning process, thereby improving the model's response speed and prediction accuracy to changes in time series data. By integrating the network architecture output of time series prediction and in-depth analysis of the nonlinear characteristics of operating data, the model calculates the deviation degree of bearing runout data, comprehensively considers the predicted values ​​within the entire observation period, and focuses on the deviation degree of operating parameters and bearing runout data. It partially highlights the impact of time attenuation, which means that the impact of data farther away from the current moment on the prediction gradually decreases. When processing nonlinear and high-dimensional operating conditions and vibration data, the model can accurately capture key features and make effective predictions, which is important for optimizing the operation management of hydropower units and providing early warning of potential operating anomalies. In addition, the dynamic adjustment mechanism of the model ensures the robustness and adaptability of the optimal base model under different operating environments and data changes, providing strong technical support for intelligent management and decision-making in the hydropower industry.

[0056] S3: Test and fine-tune the optimal foundation model, obtain the intelligent early warning model of a single unit, evaluate the performance of the hydropower unit and optimize the intelligent early warning model of a single unit.

[0057] Furthermore, obtaining a single unit intelligent early warning model includes sample testing of the optimal base model and the network architecture objective function. In and Adjust and output the optimal single unit intelligent early warning model ; When the sample size is less than or equal to 10, Adjusted to 1 for online learning, the best base model learns features from each sample.

[0058] It should be noted that the performance evaluation of the hydropower unit and optimization of the intelligent early warning model of a single unit includes the performance evaluation of the hydropower unit based on the output value M of the optimal single unit intelligent early warning model. When M is greater than 0 and less than or equal to 0.6, the hydropower station is in good operating condition and the hydropower data early warning is not triggered; when M is greater than 0.6 and less than or equal to 1, there is a problem with the operating status of the hydropower station, which triggers the hydropower data early warning and sends out an early warning signal. The early warning system prompts the operation and maintenance personnel to check the status of the unit and identify and solve the problem; by balancing the training speed and memory usage efficiency of the intelligent early warning model of a single unit, Perform dynamic adjustment and optimization, output Optimal value.

[0059] It should also be noted that, by performing sample testing and parameter adjustment on the optimal base model, the present invention realizes a highly personalized early warning model for a single unit. The intelligent early warning model for a single unit can more accurately reflect the operating status and potential risks of a specific unit, thereby providing a more accurate early warning. The present invention optimizes the speed and efficiency of model training by dynamically adjusting the number of samples and network parameters (such as batchsize). In particular, when the number of samples is small, the strategy of online learning and batchsize of 1 is adopted to ensure that the model can effectively learn even when the data is limited, which improves the ability of the model to adapt to different data environments. Through real-time evaluation of the performance of the hydropower unit and generation of early warning signals based on predetermined thresholds, the present invention can promptly discover and prompt operation and maintenance personnel to pay attention to potential operating problems, which not only helps to ensure the safe operation of the hydropower station, but also can prevent potential large-scale failures through timely maintenance and repair, thereby improving the operating efficiency and economic benefits of the hydropower station.

[0060] Example 2

[0061] An embodiment of the present invention provides a method for constructing an intelligent early warning model for hydropower data. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0062] First, we selected representative hydropower units as the research objects, ensuring that the data of the selected units can cover the operating data (such as turbine speed, flow, pressure, temperature), bearing vibration data (such as vibration frequency, amplitude), hydrological data and grid demand data. By cooperating with local hydropower stations, we obtained real-time records of these data and ensured the integrity and accuracy of the data. We defined the steady-state operation data constraint as the time point when the power fluctuation does not exceed 10MW. Such constraints help to screen out the most valuable data for model training. We designed a network architecture based on time series data prediction, set epoch and batchsize to adapt to data sets of different sizes and training requirements, and recorded and analyzed the experimental data with reference to Table 1.

[0063] Table 1 Experimental data record table

[0064]

[0065] It can be observed from the data that the speed of units A to F fluctuates around 1500RPM, while the power fluctuation remains within 10MW, which meets the constraints of the steady-state operation data. In particular, there seems to be a strong correlation between the speed and power fluctuation of unit F. When the speed is close to 1500RPM, the power fluctuation is the smallest (6.5MW), indicating that the power output of the unit is relatively stable under stable operation. Flow and pressure are key indicators of the performance of hydropower units. The data show that units with high flow (such as units C and F) seem to have lower power fluctuations, which may indicate that the units can operate more stably when there is more water flow. In addition, pressure and power are closely related. There is no direct correlation between the pressure fluctuations and the power rate fluctuations, which may be because the optimization of the unit design and operation reduces the impact of pressure fluctuations on power. The data preprocessing and time series data based model construction method proposed in the present invention can effectively identify and utilize the key features and patterns in the operation data of the hydropower unit, which not only improves the accuracy of the early warning model, but also enhances the model's adaptability to data changes under different operating conditions, thereby improving the stability and safety of the hydropower station operation. In particular, by fine-tuning the epoch and batchsize parameters, and fine-tuning and optimizing the model, the present invention shows significant advantages in improving the early warning response speed and reducing false alarms.

[0066] Example 3

[0067] Reference Figure 2 , which is an embodiment of the present invention, provides a hydropower data intelligent early warning model construction system, including a data processing module, a base construction module, and a performance evaluation module.

[0068] The data processing module is used to collect the operating data of the hydropower unit, screen the steady-state operating data and the stable operating area, and build a unit data preprocessing mechanism; the base construction module is used to set the epoch and batchsize based on the network architecture of time series data prediction, and build the optimal base model; the performance evaluation module is used to test and fine-tune the optimal base model, obtain the intelligent early warning model of a single unit, evaluate the performance of the hydropower unit, and optimize the intelligent early warning model of a single unit.

[0069] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, RandomAccess Memory), disk or optical disk, etc. Various media that can store program codes.

[0070] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0071] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk case (magnetic devices), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0072] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logical function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc. It should be noted that the above embodiments are only used to illustrate the technical solution of the present invention and are not limited. Although the present invention is described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention, which should be included in the scope of the claims of the present invention.

[0073] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for constructing an intelligent early warning model for hydropower data, characterized in that: include: Collect the operating data of hydropower units, screen steady-state operating data and stable operating areas, and build a unit data preprocessing mechanism; Based on the network architecture predicted by time series data, set epoch and batchsize to build the best base model; Test and fine-tune the optimal foundation model, obtain the intelligent early warning model of a single unit, evaluate the performance of the hydropower unit and optimize the intelligent early warning model of a single unit; Screening steady-state operation data and stable operation area includes screening steady-state operation data and stable operation area based on active power, water head and fixed properties of the unit; The steady-state operation data constraints are expressed as: AP(t,W)=maX t≤s<t+W P(s)-min t≤s<t+W P(s) AP(t,W)≤10MV Among them, AP(t, W) represents the power fluctuation range within the window size W starting at time t, and max t≤s<t+W P(s) represents the maximum power from time t to t+W, min t≤s<t+W P(s) represents the minimum power value from time t to t+W, s represents the auxiliary variable for time traversal within the window size W, and selects the time point t when the power fluctuation does not exceed 10MW within any 10-minute window; The stable operation area is a fixed interval. Within the stable operation area, the performance indicators of the unit meet the stable operation requirements of the power system. The fixed stable operation area range of each unit is screened by active power and water head. Constructing the optimal base model involves extracting features from the input data at time t in the comprehensive dataset. The feature extraction objective function F(x t , W data , b data ) is expressed as: F(x t ,W data ,b data )=tanh(W data ·x t +b data ) Among them, W data is the feature extraction weight, b data is the bias in feature extraction, x t For the input data at time t, the epoch and batch size are dynamically adjusted, and the network architecture for time series data prediction is constructed based on the feature extraction process. The network architecture objective function Net(y) is expressed as: Among them, epoch is the number of times the model is trained, batchsize represents the size of a subset selected from the comprehensive data set, α is the adjustment coefficient of epoch, β is the adjustment coefficient of batchsize, T is the length of the time series, and W combined is the weight parameter dynamically adjusted by the network architecture, b combined Dynamically adjust bias parameters for network architecture; Based on the network architecture of the operation data and time series data prediction of the hydropower unit, the nonlinear characteristics of the operating data are processed, the deviation degree of the bearing runout data is calculated, and the optimal foundation model M(x) is constructed, which is expressed as: Among them, N is the normalization factor, γ is the influence adjustment coefficient of the network architecture objective function, K is the total number of parameters in the working condition data, and w k is the weight of the kth parameter in the operating condition data, D wk represents the value of the kth operating condition parameter, L is the total number of parameters in the bearing runout data, and D bl represents the observed value of the first bearing runout data, Indicates the average value of the first bearing runout data.

2. The method for constructing a hydropower data intelligent early warning model according to claim 1, characterized in that: The operation data of the hydroelectric unit includes operating condition data, bearing runout data, hydrological data and power grid demand data; The operating data include turbine speed, flow, pressure and temperature; Bearing runout data includes vibration frequency and amplitude.

3. The method for constructing a hydropower data intelligent early warning model according to claim 1, characterized in that: The unit data preprocessing mechanism includes data standardization, abnormal data screening and data splicing of the collected hydropower unit operation data; Abnormal data screening includes marking the values ​​lasting more than 3 minutes as dead values, screening out the dead values, and standardizing the operating data of the hydropower unit, which is expressed as: Among them, Z i Represented as the i-th standardized data, X i is the ith original data, μ and σ are the mean and standard deviation of the original data respectively, and the operating characteristic weight W of the hydropower unit is introduced i , identify and filter outliers, expressed as: in, is the optimized local anomaly factor value, x is any observation point in the data set, N k (x i ) is the i-th data point x i The k nearest neighbor set, lrdk(x i ) is the i-th data point x i The local reachable density of the i-th data point x i The density of the surrounding area, lrd k (x) is the local reachable density of any observation point. When the i-th data point x i Mark as outliers and screen out; The operating data of the hydropower units are stitched into a comprehensive data set.

4. The method for constructing a hydropower data intelligent early warning model according to claim 1, characterized in that: The obtaining of the single unit intelligent early warning model includes performing sample testing on the optimal base model, adjusting the epoch and batchsize in the network architecture objective function Net(y), and outputting the optimal single unit intelligent early warning model M′(x); When the number of samples is less than or equal to 10, the batch size is adjusted to 1 for online learning, and the best base model learns features from each sample.

5. The method for constructing a hydropower data intelligent early warning model according to claim 4, characterized in that: The performance evaluation of the hydropower unit and optimization of the single unit intelligent early warning model includes performing a performance evaluation of the hydropower unit based on the output value M of the optimal single unit intelligent early warning model, when M is greater than 0 and less than or equal to 0.6, the hydropower station is in good operating condition and does not trigger a hydropower data early warning; When M is greater than 0.6 and less than or equal to 1, there is a problem with the operation status of the hydropower station, which triggers the hydropower data warning and issues a warning signal. The warning system prompts the operation and maintenance personnel to check the status of the unit and identify and solve the problem; By balancing the training speed and memory usage efficiency of the intelligent early warning model of a single unit, the batch size is dynamically adjusted and optimized to output the optimal batch size value.

6. A system using the method for constructing a hydropower data intelligent early warning model according to any one of claims 1 to 5, characterized in that: Including data processing module, base construction module, performance evaluation module; The data processing module is used to collect the operating data of the hydropower unit, screen the steady-state operating data and the stable operating area, and build a unit data preprocessing mechanism; The base building module is used to set the epoch and batch size based on the network architecture predicted by time series data and build the best base model; The performance evaluation module is used to test and fine-tune the optimal foundation model, obtain a single unit intelligent early warning model, perform performance evaluation on the hydropower unit and optimize the single unit intelligent early warning model.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the steps of the method for constructing a hydropower data intelligent early warning model according to any one of claims 1 to 5 are implemented.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for constructing a hydropower data intelligent early warning model according to any one of claims 1 to 5 are implemented.