Hydroelectric generating set state prediction system based on deep learning

Abstract

The invention discloses a hydroelectric generating set state prediction system based on deep learning, and relates to the technical field of hydroelectric machine monitoring, and the system comprises the steps: collecting a hydroelectric generating set through a data sensor network, and obtaining a plurality of operation sensing data; carrying out deep learning on the plurality of operation time periods, and constructing a state prediction channel; synchronizing the plurality of operation sensing data to a state prediction channel for analysis, generating a plurality of operation state prediction results, performing operation monitoring on the hydroelectric generating set, and generating an operation monitoring result; traversing a plurality of operation state prediction results for verification, generating prediction feedback information, and updating and optimizing a state prediction channel; and the state of the hydroelectric generating set is intelligently predicted. According to the method, the technical problem of low reliability and operation efficiency of the hydroelectric generating set caused by inaccurate prediction of the operation state of the hydroelectric motor in the prior art is solved, and the technical effects of realizing accurate and intelligent prediction of the operation state of the hydroelectric generating set and improving the reliability and operation efficiency of the hydroelectric generating set are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of hydroelectric generator monitoring, and in particular to a hydroelectric generator set state prediction system based on deep learning. Background Art

[0002] In today's power production field, hydropower units, as an important clean energy power generation equipment, play a key role. However, traditional hydropower unit status monitoring and prediction methods have certain limitations. On the one hand, traditional methods usually rely on data collection and simple data analysis of a single sensor, which is difficult to fully and accurately reflect the complex operating status of hydropower units. Moreover, for the changes in the operating status of hydropower units in multiple operating periods and different operating cycles, traditional methods often lack effective processing methods and are difficult to capture the dynamic change characteristics in time series. On the other hand, traditional status prediction mainly relies on empirical models and fixed rules, lacks the ability to deeply mine and learn a large amount of historical operating data, and has limited prediction accuracy and adaptability. At the same time, when faced with abnormal operating conditions, traditional methods have insufficient diagnostic and positioning capabilities, making it difficult to quickly determine the source of abnormal data, resulting in inefficient troubleshooting and maintenance.

[0003] The existing technology has the technical problem of inaccurate prediction of the operating status of the hydropower generator, resulting in low reliability and operating efficiency of the hydropower generator unit. Summary of the invention

[0004] The present application provides a hydropower unit state prediction system based on deep learning, which is used to solve the technical problem in the prior art of inaccurate prediction of the operating state of hydropower units, resulting in low reliability and operating efficiency of hydropower units.

[0005] In view of the above problems, the present application provides a hydropower unit state prediction system based on deep learning.

[0006] The present application provides a hydropower unit state prediction system based on deep learning, the system comprising:

[0007] An operation sensor data acquisition module, the operation sensor data acquisition module collects the hydropower unit in real time through the data sensor network based on multiple operation time periods to obtain multiple operation sensor data; a state prediction channel construction module, the state prediction channel construction module is used to perform deep learning according to the historical operation sensor data set according to the multiple operation time periods to construct a state prediction channel; an operation monitoring result generation module, the operation monitoring result generation module is used to synchronize the multiple operation sensor data to the state prediction channel for analysis, generate multiple operation state prediction results, perform operation monitoring on the hydropower unit according to the multiple operation state prediction results combined with the operation environment data set, and generate operation monitoring results, the operation monitoring results include multiple state tags; a state prediction optimization channel generation module, the state prediction optimization channel generation module is used to traverse the multiple operation state prediction results according to the multiple state tags for verification, generate prediction feedback information, update and optimize the state prediction channel according to the prediction feedback information, and generate a state prediction optimization channel; an intelligent prediction module, the intelligent prediction module is used to perform intelligent prediction of the state of the hydropower unit through the state prediction optimization channel.

[0008] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0009] The operation sensor data acquisition module collects the hydropower unit in real time through the data sensor network based on multiple operation periods to obtain multiple operation sensor data; the state prediction channel construction module performs deep learning according to the historical operation sensor data set according to the multiple operation periods to construct the state prediction channel; the operation monitoring result generation module synchronizes the multiple operation sensor data to the state prediction channel for analysis, generates multiple operation state prediction results, and performs operation monitoring on the hydropower unit according to the multiple operation state prediction results combined with the operation environment data set to generate operation monitoring results; the state prediction optimization channel generation module traverses the multiple operation state prediction results according to the multiple state labels for verification, generates prediction feedback information, updates and optimizes the state prediction channel, and generates a state prediction optimization channel; the intelligent prediction module performs intelligent prediction of the state of the hydropower unit through the state prediction optimization channel. The technical effect of realizing accurate and intelligent prediction of the operation state of the hydropower unit and improving the reliability and operation efficiency of the hydropower unit is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0011] Figure 1 A structural schematic diagram of a hydropower unit state prediction system based on deep learning is provided for an embodiment of the present application;

[0012] Figure 2 A schematic diagram of the output state prediction channel flow of a hydropower unit state prediction system based on deep learning is provided for an embodiment of the present application.

[0013] Explanation of the reference numerals: operation sensor data acquisition module 10, state prediction channel construction module 20, operation monitoring result generation module 30, state prediction optimization channel generation module 40, intelligent prediction module 50. DETAILED DESCRIPTION

[0014] The present application provides a hydropower unit state prediction system based on deep learning, which is used to solve the technical problem in the prior art of inaccurate prediction of the operating state of hydropower units, resulting in low reliability and operating efficiency of hydropower units.

[0015] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0016] Embodiment 1, as Figure 1 As shown, the present application provides a hydropower unit state prediction system based on deep learning, the system comprising:

[0017] The operation sensor data acquisition module 10 collects data from the hydropower unit in real time through a data sensor network based on multiple operation time periods to obtain multiple operation sensor data.

[0018] Specifically, the operation sensor data acquisition module 10 focuses on acquiring the operation sensor data of the hydropower unit, and its working process is based on multiple operation periods. In each operation period, the data sensor network will conduct comprehensive and real-time monitoring of the hydropower unit. This data sensor network is a complex network composed of multiple types of sensors, which are carefully deployed in various key parts of the hydropower unit. For example, temperature sensors are installed on the runner, bearing, volute and other parts of the hydropower unit to monitor the temperature changes of the components; pressure sensors are set at the pipe connection and the pressure chamber to obtain pressure data; vibration sensors are equipped near the support structure and key transmission parts of the unit to detect information such as vibration frequency and amplitude. Through these widely distributed sensors with different functions, a large amount of raw data about the operation status of the hydropower unit can be obtained in each operation period. These raw data cover all aspects of the unit operation, including but not limited to parameters such as temperature, pressure, vibration, and flow. As time goes by, under the continuous collection of multiple operation periods, multiple operation sensor data can be obtained. These data form a rich data set, which provides a basis for subsequent analysis and processing, and is an important prerequisite for the accurate operation of the entire hydropower unit state prediction system.

[0019] The state prediction channel construction module 20 is used to perform deep learning according to the historical operation sensor data set according to the multiple operation time periods to construct a state prediction channel.

[0020] Specifically, the state prediction channel construction module 20 is a key part of the entire hydropower unit state prediction system, and is responsible for constructing a channel for predicting the state of the hydropower unit. First, a historical operation sensor data set is obtained. This data set contains various operation parameter records of the hydropower unit in multiple operation periods in the past. These records are a comprehensive reflection of the unit's operating state. For example, it covers information such as temperature changes, pressure fluctuations, vibration conditions, and flow data in different periods. Then, deep learning is performed on the historical operation sensor data set according to multiple operation periods. Before deep learning, the historical operation sensor data set needs to be preprocessed according to multiple operation periods, which includes cleaning the data, removing outliers and erroneous data, and standardizing the data so that data with different characteristics have similar scales for better learning. The preprocessed historical operation sensor data is input into the LSTM network in time series. For each time step, the data enters the memory unit through the input gate. Inside the memory unit, the forget gate determines which information needs to be forgotten based on the current input and the state of the previous moment. For example, if a parameter change in a certain operation period has little effect on the subsequent state, the forget gate may weaken or forget the relevant information. At the same time, the input gate determines which new information needs to be stored in the memory unit. For example, if a new operating feature appears in the current period and has an important impact on subsequent predictions, the input gate will include it in the memory unit. After being processed by the forget gate and the input gate, the information in the memory unit is output through the output gate. The output information will be used as part of the input of the next time step, and will also be part of the output result of the entire network for the current time step. This output result will gradually reflect the patterns and laws in the historical operation sensor data as time accumulates and continuously learns. Through long-term learning and processing of a large amount of historical operation sensor data, the LSTM network can learn the complex relationship and dynamic change pattern between hydropower units in different operating periods. Based on these learned patterns, a state prediction channel that can predict the future state of the hydropower unit is constructed. When new operation sensor data is input, this channel can use the previously learned knowledge to predict the future state of the hydropower unit.

[0021] An operation monitoring result generating module 30 is used to synchronize the multiple operation sensor data to the state prediction channel for analysis, generate multiple operation state prediction results, perform operation monitoring on the hydropower unit according to the multiple operation state prediction results combined with the operation environment data set, and generate operation monitoring results, wherein the operation monitoring results include multiple state tags.

[0022] Specifically, the operation monitoring result generation module 30 synchronizes the multiple operation sensor data obtained from the operation sensor data acquisition module to the state prediction channel. This synchronization process ensures that the data can be analyzed in the constructed state prediction channel. The state prediction channel is constructed based on the LSTM network algorithm to learn the historical operation sensor data set, and has the ability to predict the operation state of the hydropower unit. When the operation sensor data enters the channel, it is processed by the algorithm and model in the channel to generate multiple operation state prediction results. These results reflect the possible operation state of the hydropower unit in the future, such as predicting that the temperature of a certain component will rise, or the pressure of a certain part will fluctuate. Then, according to these operation state prediction results, the operation monitoring of the hydropower unit is carried out in combination with the operation environment data set. The operation environment data set contains environmental factors such as water temperature, water level, water quality, etc. related to the operation of the hydropower unit. By comprehensively considering the prediction results and environmental factors, the actual operation of the hydropower unit can be more comprehensively understood. For example, if it is predicted that the temperature of a certain component will rise, combined with environmental factors such as the current water temperature and water flow rate, it is more accurate to judge whether the component is in an abnormal state. Finally, based on the above comprehensive analysis, the operation monitoring results are generated, which contain multiple status tags, which are used to describe whether the operation status of the hydropower unit is normal, whether there are abnormalities, and the type of abnormalities. For example, there will be different status tags such as "normal operation", "temperature abnormality", and "pressure fluctuation abnormality", which can provide important basis for subsequent maintenance and decision-making.

[0023] The state prediction optimization channel generation module 40 is used to traverse the multiple operating state prediction results according to the multiple state tags for verification, generate prediction feedback information, update and optimize the state prediction channel according to the prediction feedback information, and generate a state prediction optimization channel.

[0024] Specifically, the state prediction optimization channel generation module 40 is responsible for optimizing the state prediction channel to improve the accuracy and reliability of the prediction. First, according to the multiple state labels generated by the operation monitoring result generation module, multiple operation state prediction results are traversed and verified, and for each state label, the corresponding operation state prediction result is checked to see whether it is accurate. For example, if the state label shows that a certain component has "temperature abnormality", then it will be checked whether the temperature abnormality of the component is correctly predicted in the prediction result. In this traversal verification process, the module will analyze in detail the deviation between the prediction result and the actual operation state. According to the deviation found in the above verification process, the module generates prediction feedback information, which contains various details about the accuracy of the prediction, such as which prediction results are inaccurate, what is the specific value of the deviation, what is the reason for the deviation, etc. For example, if it is found that the temperature prediction value of a certain component deviates greatly from the actual value, the feedback information will clearly point out the component, the deviation value and the factors that may cause the deviation, such as sensor failure, environmental factors not fully considered, etc. Finally, the state prediction channel is updated and optimized according to the generated prediction feedback information. The module adjusts various parameters and algorithm structures in the state prediction channel. For example, if it is found that a certain type of data has a greater impact on the prediction results, but the weight setting in the channel is unreasonable, then its weight will be readjusted; if it is found that a certain algorithm does not work well when dealing with a specific situation, an attempt will be made to replace or improve the algorithm. Through these update optimization measures, a state prediction optimization channel is eventually generated, enabling it to more accurately predict the state of the hydropower unit.

[0025] The intelligent prediction module 50 is used to perform intelligent prediction of the state of the hydropower unit through the state prediction optimization channel.

[0026] Specifically, the intelligent prediction module 50 uses the state prediction optimization channel to perform intelligent prediction of the state of the hydropower unit. When the current operating data of the hydropower unit is received, the intelligent prediction module 50 will input these data into the state prediction optimization channel. The state prediction optimization channel is a prediction model optimized by the previous module. It contains the patterns and rules learned from the historical operating sensor data, as well as the algorithm and parameter settings after the prediction inaccuracy is corrected. Inside the state prediction optimization channel, the input operating data will be processed according to the algorithm and rules set by the channel. Through these processing processes, the intelligent prediction module 50 finally outputs the future operating state prediction results of the hydropower unit. These prediction results are predictions of temperature changes of a certain component, predictions of pressure changes, predictions of vibration conditions, etc. These predictions can help understand the operating trend of the unit in advance so as to take corresponding maintenance measures, thereby ensuring the safe and stable operation of the hydropower unit.

[0027] In one possible implementation, Figure 2 As shown, the state prediction channel building module also includes:

[0028] Based on the multiple operating time periods, an operation analysis of the hydropower unit is performed, and a multiple operation cycle is generated according to the analysis results; the historical operation data archive of the hydropower unit is retrieved, the historical operation sensor data set is extracted according to the historical operation data archive, and the feature extraction is performed on the historical operation sensor data set according to the multiple operation cycles to generate a multiple historical operation features; a time series analysis is performed according to the multiple operation cycles combined with the multiple historical operation features to determine the operation time step of the hydropower unit; a long short-term memory network is used to perform deep learning on the historical operation sensor data set according to the operation time step to generate an operation time series, wherein the operation time series includes multiple operation time nodes, and each of the multiple operation time nodes has at least one historical operation sensor data; the multiple historical operation sensor data are verified and evaluated according to the multiple operation time nodes to generate a verification token, and an evaluation is performed according to the verification token to output a state prediction channel.

[0029] Specifically, first, it is necessary to collect detailed data of the hydropower unit in each operating period, including but not limited to power output, water flow, unit speed, temperature change, pressure fluctuation and other parameters. These data can fully reflect the operating status of the unit in different periods. Then, a comprehensive analysis of these data is conducted. For example, by observing the curve of power output over time, it may be found that the power output of the unit is relatively stable in some periods, while it shows a clear upward or downward trend in other periods. For water flow, it may be found that there is a certain correlation between it and power output. When the power output is high, the water flow is also relatively large. At the same time, temperature changes and pressure fluctuations will also affect the operation of the unit. Excessive temperature or unstable pressure may indicate a change in the operating state of the unit. Through detailed analysis and comparison of these data, according to their changing laws and characteristics, the operation process of the hydropower unit is divided into multiple operation cycles. Each operation cycle has unique operating characteristics. For example, in a certain operation cycle, the unit is in a stable power generation state, and parameters such as power output, water flow, temperature and pressure are relatively stable; while in another operation cycle, the unit is in the process of starting or shutting down, and these parameters will change greatly. This division method helps to gain a deeper understanding of the operating mechanism of hydropower units and provides a more reasonable framework for subsequent data analysis and processing.

[0030] Retrieving the historical operation data archive of the hydropower unit is a key step in obtaining data. This archive stores a large amount of data records of the unit during its long-term operation in the past. These records cover various aspects of operation information. The historical operation sensor data set is extracted from the historical operation data archive. This data set is obtained after screening and sorting the original archive data. It mainly contains the unit operation parameter data collected by various sensors, such as the temperature value recorded by the temperature sensor, the pressure value recorded by the pressure sensor, the flow value recorded by the flow sensor, and the vibration parameter recorded by the vibration sensor. These data can directly reflect the operation status of the unit at different times. Next, the historical operation sensor data set is feature extracted according to the multiple operation cycles divided previously. In each operation cycle, each parameter in the data set is analyzed to extract features that can represent the operation characteristics of the cycle. For temperature data, the average temperature, maximum temperature, minimum temperature, and temperature fluctuation range in the cycle are calculated. These features reflect the temperature changes of the unit in the cycle, such as whether the unit is in a stable temperature environment, or whether there is an abnormal situation of too high or too low temperature. For pressure data, calculate the average pressure, maximum pressure, minimum pressure, and pressure change rate. These pressure characteristics can help determine whether the unit is under stable pressure during the cycle and whether the pressure change is within the normal range. For flow data, calculate the average flow, maximum flow, minimum flow, and flow change trend. The flow characteristics reflect the water supply of the unit during the cycle, whether it meets the power generation demand, and whether the flow is stable. By performing the above feature extraction operations on different parameters in each operating cycle, multiple historical operating characteristics are finally generated. These historical operating characteristics can more concisely and accurately describe the operating status and characteristics of the hydropower unit in each operating cycle, providing a more targeted data basis for subsequent time series analysis and deep learning.

[0031] Consider multiple operation cycles, each of which represents the operation mode of the hydropower unit at different stages, and these modes differ in power output, temperature change, pressure fluctuation, flow adjustment, etc. For example, one operation cycle corresponds to the startup phase of the unit, when the power gradually increases, and the temperature and pressure also change accordingly; another operation cycle is the stable operation phase, and various parameters are relatively stable. At the same time, combined with multiple historical operation features, these features are extracted from the historical operation sensor data set of each operation cycle, which can describe the characteristics of each operation cycle in more detail. For example, in a certain operation cycle, features such as average temperature, temperature fluctuation range, and pressure change rate reflect the operating status of the unit in that cycle. When performing time series analysis, it is necessary to study the change rules of these operation cycles and their corresponding historical operation features on the time axis, observe the transition between different operation cycles, and the change trend of features over time in each cycle. For example, analyze how the power gradually stabilizes from the startup phase to the stable operation phase, how the temperature and pressure reach stable values, and how long these changes take to complete. Through the time series analysis of these operating cycles and historical operating characteristics, the operating time step of the hydropower unit is determined. This time step is an appropriate time interval that can accurately capture the key changes in the unit's operation process, and will not miss important information due to a large time interval, nor will it cause the data to be too complicated and difficult to process due to a small time interval. For example, if the time step is set to 1 minute, then when analyzing the data, the operating status changes of the unit can be observed every 1 minute, and subsequent deep learning and other operations can be performed based on this.

[0032] The historical operating sensor data set is processed according to the set operating time step. This operating time step determines the time interval for data to enter the network, so that the data can be received and processed by the LSTM network in an orderly manner. For example, if the operating time step is 5 minutes, then every 5-minute data fragment will be input into the network as a unit. Then, when the data unit enters the LSTM network, the network begins the learning process. The unique structure of the LSTM network includes an input gate, a forget gate, and an output gate. The forget gate determines which information to forget based on the current input and the state of the previous moment. For example, if a sensor data from a previous time period is not important for the current prediction, the forget gate will reduce its weight in the memory. At the same time, the input gate will determine which new information to store in the memory unit based on the current input. For example, if there is a new pressure fluctuation feature in the current time period, the input gate will include the relevant data in the memory unit. Inside the network, the information processed by the forget gate and the input gate is updated in the memory unit and output through the output gate. The output information not only affects the input of the next time step, but also contributes to the output of the current time step. In this way, the LSTM network can capture long-term dependencies and dynamic change patterns in the data. As time goes by and data is continuously input, the network eventually generates an operation time series. This operation time series contains multiple operation time nodes, each of which corresponds to a specific time point (determined by the operation time step). For example, there is a node every 5 minutes, and each operation time node has at least one historical operation sensor data. These historical operation sensor data are processed and integrated by the network, reflecting the operation status of the hydropower unit at the corresponding time point and its relationship with other time points, providing an important basis for subsequent analysis and prediction.

[0033] For each operation time node and the corresponding historical operation sensor data, the rationality of the time period must be checked first. Considering the operation process and logic of the hydropower unit, different operation time periods have specific physical processes and characteristics. For example, during the unit startup phase, some data related to stable operation may not appear or be in an unstable state. At this time, it is necessary to determine whether there are some data that do not conform to the characteristics of the current time period. For example, if precise temperature control data that can only be obtained under stable operation appears in the early stage of startup, this is unreasonable because the unit is still in the process of heating up and has not yet reached the stable temperature control stage. Similarly, during the shutdown phase, if high power output data during normal operation appears, it is obviously not in line with the actual situation. It is also necessary to check whether there are some data that cannot exist in the current time period from a physical principle. For example, when the unit is in a static and unstarted state, the flow sensor should not detect water flow data. If such data appears, it is likely that the sensor is faulty or the data is recorded incorrectly. This check on the rationality of the time period is an important step in ensuring data quality, which helps to exclude abnormal data caused by incorrect collection or non-compliance with the operation logic. In addition to the rationality check of the time period, it is also necessary to check the data range of each sensor data. The physical quantities measured by different sensors have their reasonable value ranges, which depend on the design and operating characteristics of the hydropower unit. For example, the temperature range usually measured by the temperature sensor should be within the temperature range that the unit can withstand. If data outside this range appears, it means that the sensor is faulty or the unit has an abnormal condition, such as overheating or overcooling. The pressure data measured by the pressure sensor should also be within a reasonable pressure range. If abnormal high or low pressure data appears, the cause also needs to be further investigated. By checking the data range, the data with problems can be further screened out.

[0034] Based on the results of the above data verification and evaluation, the evaluation indicators used to generate the verification token are determined. These evaluation indicators can quantify the quality of the data and provide a basis for the subsequent generation of verification tokens. By comprehensively considering these indicators, a more comprehensive understanding of the overall quality of the data can be obtained. Based on the determined evaluation indicators, a verification token is generated. The verification token is a numerical value, such as a score between 0 and 1. The higher the score, the better the data quality, reflecting the overall situation of the data after verification and evaluation.

[0035] The generated verification token is evaluated. If the verification token meets the pre-set standards, for example, the numerical verification token is greater than a certain threshold (such as 0.8), or each indicator in the data object verification token meets the requirements (such as the proportion of data that does not meet the rationality of the time period is lower than a certain value, and the proportion of data that exceeds the data range is lower than a certain value), it means that the data quality has met the requirements and the state prediction channel is output. If the verification token does not meet the standards, it is necessary to recheck the data collection process to see if there are problems such as sensor failure and data recording errors. After repairing the problems found, the verification evaluation is performed again until the verification token meets the requirements. When the verification token meets the requirements, the output state prediction channel is a verified and optimized model structure that can predict the future state of the hydropower unit based on the relevant input data.

[0036] In a possible implementation, the state prediction channel construction module further includes:

[0037] According to the operation time step, the historical operation sensor data set is subjected to data time series standardization processing to generate a historical sensor standard data set; the historical sensor standard data set is segmented according to the operation time step to generate a plurality of standard data sample segments; based on the plurality of standard data sample segments, a long-term analysis is performed to generate a first analysis result, a plurality of long-term standard data sample segments are extracted according to the first analysis result, and calculations are performed according to the plurality of long-term standard data sample segments to generate a long-term dependency coefficient; based on the plurality of standard data sample segments, a short-term analysis is performed to generate a second analysis result, a plurality of short-term standard data sample segments are extracted according to the second analysis result, and calculations are performed according to the plurality of short-term standard data sample segments to generate a short-term dependency coefficient; the historical operation sensor data set is captured according to the long-term dependency coefficient to generate a first information flow; the historical operation sensor data set is captured according to the short-term dependency coefficient to generate a second information flow; the first information flow and the second information flow are stacked to determine a plurality of operation time nodes, and the operation time series is extracted by integrating the plurality of operation time nodes.

[0038] Specifically, when performing data time series standardization on the historical operation sensor data set, the Z-score standardization algorithm is used to divide the historical operation sensor data set into different time segments according to the operation time step to ensure that the data of each segment is coherent in time. For the data in each time segment, its mean and standard deviation are calculated. The mean reflects the central trend of the data in the time period, and the standard deviation reflects the degree of dispersion of the data. For each data point in the data set, the standardization is performed by the following formula: Standardized value = (original value - mean of the time period) / standard deviation of the time period. For example, suppose there is a temperature sensor that collects data at different time points. In a specific operation time step, the mean of the temperature data is 25°C and the standard deviation is 2°C. If the temperature value at a certain moment is 28°C, then after standardization, the value becomes (28-25) / 2=1.5. Through such processing, the historical operation sensor data set is converted into a historical sensor standard data set.

[0039] The historical sensor standard data set is segmented according to the running time step to generate multiple standard data sample segments. Each standard data sample segment corresponds to a specific time interval, and its length is determined by the running time step. This segmentation method helps to analyze the data more carefully, breaking down a large data set into multiple smaller data units with similar time scales, providing a basis for subsequent long-term and short-term analysis.

[0040] Long-term analysis is performed on multiple standard data sample segments. This analysis process uses statistical analysis and trend analysis methods to observe the changing patterns and characteristics of these data over a long time span. After analysis, the first analysis result will be generated, which contains information such as the long-term changing trend of certain parameters and the long-term correlation between different parameters. Then, based on the first analysis result, multiple long-term standard data sample segments that are representative and important are extracted. These sample segments are the data parts that show obvious trends or special patterns in long-term analysis. Then, the covariance and correlation coefficients between different parameters are calculated for these multiple long-term standard data sample segments, and finally a long-term dependence coefficient is generated. This long-term dependence coefficient can quantify the dependence between different parameters of the hydropower unit during long-term operation, and provide an important basis for evaluating the long-term stability of the hydropower unit and predicting its future operating status.

[0041] Perform short-term analysis on multiple standard data sample segments. In this process, we will focus on the data changes within a shorter time range, analyze the fluctuations of the operating parameters of the hydropower unit within a day, a few hours or even a shorter time, and through detailed research on these data, use some specific analysis methods to calculate the mean, variance, change rate and other statistics of the data in the short term. After the short-term analysis, the second analysis result will be generated. This result includes some obvious features in the short term, such as the rapid changes of certain parameters in a specific time period, the short-term strong correlation between parameters, etc. Based on the second analysis results, extract multiple short-term standard data sample segments. These sample segments are the data parts that are considered to have important characteristics or representativeness in the short-term analysis. For example, if it is found that the vibration parameters of the hydropower unit fluctuate abnormally within a certain hour, then the relevant data sample segments of this time period may be extracted for further analysis. Then, calculate based on these multiple short-term standard data sample segments to generate a short-term dependence coefficient. The calculation of this coefficient is similar to the calculation method of the long-term dependence coefficient, but it focuses more on the short-term time scale. By calculating the covariance of different parameters in the short term, their dependence in the short term can be measured. The short-term dependence coefficient reflects the characteristics of the hydropower unit's operating state changes in the short term and the degree of mutual influence between parameters. It can help quickly understand the hydropower unit's operating stability and potential problems in the short term, and provide a basis for timely adjustment measures or fault warnings.

[0042] The historical operation sensor data set is captured according to the long-term dependency coefficient, which reflects the relationship between the operating parameters of the hydropower unit on a long-term time scale. Using this coefficient, data points with important correlations in the long term are screened out from the historical operation sensor data set. For example, if the long-term dependency coefficient indicates that temperature and pressure have a strong positive correlation in the long term, then during the capture process, the focus will be on those data points where temperature and pressure show a specific change trend at the same time. Through this targeted capture of the historical operation sensor data set, the first information flow is generated, which contains key information about the hydropower unit on a long-term time scale, such as long-term stable operation modes, slowly changing trends, and parameter combinations related to long-term performance.

[0043] Similarly, the historical operation sensor data set is captured according to the short-term dependency coefficient, which reflects the mutual relationship and dynamic changes of the operating parameters of the hydropower unit in the short term. Using this coefficient, the data points with significant correlation in the historical operation sensor data set in the short term are quickly identified. For example, if the short-term dependency coefficient shows that the vibration frequency and power output are closely related in a short period of time, then in the capture process, the data points where the vibration frequency and power output change simultaneously will be selected. After such a capture operation, the second information stream is generated. The second information stream covers the key information of the hydropower unit on the short-term time scale, such as sudden changes, short-term fluctuations, and fast-response parameter combinations, which is very important for timely detection of short-term abnormal conditions of the hydropower unit and rapid adjustment of the operation strategy.

[0044] The first information flow and the second information flow are stacked. This stacking process integrates long-term and short-term information to fully understand the operating status of the hydropower unit at different time scales. Through stacking, the relationship between long-term trends and short-term fluctuations, as well as their combined impact on the operation of the hydropower unit, is discovered. Based on the stacked information flow, multiple operating time nodes are determined. These time nodes are determined based on key change points, turning points, or data points of great significance in the information flow. For example, the moment when the long-term trend changes significantly, the moment when the short-term fluctuation reaches a peak or valley, etc. Finally, these multiple operating time nodes are integrated to extract the operating time series. This operating time series integrates long-term and short-term information and reflects the changes in the operating status of the hydropower unit at different time points. It provides a comprehensive and accurate time series data basis for subsequent analysis, prediction, and decision-making.

[0045] In a possible implementation, the state prediction channel construction module further includes:

[0046] The first information flow and the second information flow are aligned according to the running time step to generate a timestamp alignment result; based on the timestamp alignment result, the first information flow and the second information flow are stacked according to the time dimension to generate a two-dimensional data matrix; the two-dimensional data matrix is ​​traversed to perform information change analysis to generate an information change curve, and the multiple running time nodes are extracted according to the information change curve; multi-dimensional integration is performed according to the multiple running time nodes to generate a multi-dimensional vector set, and the multi-dimensional vector set is arranged according to the multiple running cycles to determine the running time series.

[0047] Specifically, the first information stream is aligned with the second information stream according to the running time step. The running time step is used as a key time scale to ensure that the two information streams can accurately correspond in time. In this process, each data point is compared and adjusted according to the timestamp, so that data points with the same or similar timestamps can be accurately matched to generate a timestamp alignment result. Such an alignment operation enables subsequent analysis to be performed in a unified time frame, avoiding erroneous analysis due to time inconsistency.

[0048] Based on the timestamp alignment results, stacking the first information stream and the second information stream according to the time dimension to generate a two-dimensional data matrix is ​​a key data processing step. First, the timestamp alignment results ensure the temporal correspondence of the two information streams, providing an accurate basis for the subsequent stacking operation. Then, in the stacking process, the time step is used as the row of the two-dimensional data matrix, which means that each row represents a specific time point, and the columns represent different dimensions, which include various data features from the first information stream and the second information stream. For example, one column represents a specific parameter in the first information stream, and the other column represents another parameter in the second information stream. In this way, the two information streams are integrated into a two-dimensional data matrix, so that the data changes in different dimensions can be observed simultaneously in the same time frame. Such a two-dimensional data matrix provides an intuitive and structured data set for further data analysis, which is convenient for various statistical analysis, pattern recognition and other operations, and helps to deeply understand the operating status of the hydropower unit at different time points and in different aspects.

[0049] When processing the two-dimensional data matrix, the information change analysis is performed by traversing the matrix. During the traversal process, the data value changes of different dimensions corresponding to each time step are carefully observed. For example, the numerical differences of each data dimension in adjacent time steps are compared, and the data change rate, fluctuation amplitude and other indicators are calculated to understand the dynamic change characteristics of the data over time. Through a comprehensive analysis of the entire two-dimensional data matrix, an information change curve is generated. This curve intuitively shows the change trend and amplitude of the data at different time points. According to the information change curve, multiple operating time nodes are extracted. In this process, the time points with large data fluctuations are focused on, because these time points often mean that the operating status of the hydropower unit has changed significantly. For example, if the values ​​of multiple data dimensions have changed significantly at a certain time point, then this time point may be extracted as an operating time node. At the same time, nodes with specific state changes will also be selected, such as the moment when the stable operating state changes to the fault state, or the time point when one operating mode switches to another operating mode. These extracted operating time nodes are of great significance for understanding the operating dynamics and key events of the hydropower unit, and provide key time references for subsequent analysis and decision-making.

[0050] Multi-dimensional integration is performed based on multiple operating time nodes. For each operating time node, all the data of all dimensions involved, such as temperature, pressure, vibration, flow, etc., are integrated. These data of different dimensions are combined into a unified vector, so that multiple information on a time node can be integrated. For example, at a specific operating time node, the temperature value, pressure value, vibration amplitude value and flow value at that moment are arranged in a certain order into a vector, and this vector fully represents the operating status characteristics of the hydropower unit at that time node. By performing such processing on each operating time node, a multi-dimensional vector set is generated. This multi-dimensional vector set is arranged according to multiple operating cycles, and the operating cycle is a time interval determined according to the actual operating conditions of the hydropower unit. In the arrangement process, ensure that each multi-dimensional vector corresponds to the correct time position, and arrange these vectors in order according to the chronological order. In this way, the operating time sequence can be determined by orderly arranging the multi-dimensional vector set. This operation time series not only contains the detailed status information of the hydropower unit at different time points, but also organically combines data of different dimensions in the form of multi-dimensional vectors, providing a comprehensive and accurate time series data basis for further analysis of the operation trend of the hydropower unit, fault diagnosis and prediction, etc.

[0051] In a possible implementation, the state prediction channel construction module further includes:

[0052] The multiple operation sensor data are synchronized to the state prediction channel and stored in a memory module of the state prediction channel, wherein the memory module includes the historical operation sensor data set; the multiple operation sensor data are distance calculated with the historical operation sensor data set through the memory module to determine multiple Euclidean distance data; it is judged whether the multiple Euclidean distance data are in an expected distance interval, and the Euclidean distance data that are not in the expected distance interval are extracted to determine multiple abnormal distances; the historical operation sensor data set is traversed using the multiple abnormal distances as indexes to generate multiple sensor data to be updated; according to the multiple operation cycles, the multiple operation sensor data are mapped and replaced with the multiple sensor data to be updated, and the memory module is updated to obtain a memory update module; regression prediction is performed based on the memory update module and the multiple operation sensor data sets to generate the multiple operation state prediction results.

[0053] Specifically, multiple operation sensor data are synchronously transmitted to the state prediction channel, and these data are stored in the memory module of the state prediction channel. This memory module already contains the historical operation sensor data set. The addition of new operation sensor data further enriches the data collection. In this way, the integration of new and old data in the same module is achieved, providing a more comprehensive data basis for subsequent analysis and processing.

[0054] By utilizing the function of the memory module, the distance between the newly stored multiple operation sensor data and the historical operation sensor data set is calculated. The Euclidean distance calculation method is used here. By calculating each corresponding data point, multiple Euclidean distance data can be determined. These distance data reflect the degree of difference between the new operation sensor data and the historical data in numerical value, which can help quantify the similarity or difference between the two.

[0055] These multiple Euclidean distance data are judged to see whether they are in the expected distance interval. The expected distance interval is a range pre-set according to the normal operation of the hydropower unit and the statistical characteristics of the data. If a certain Euclidean distance data is not in this expected distance interval, it means that the new operation sensor data corresponding to the data is significantly different from the historical data, and there is an abnormality. The Euclidean distance data that are not in the expected distance interval are extracted and determined as multiple abnormal distances.

[0056] Using these multiple abnormal distances as indexes, the historical running sensor data set is traversed. During the traversal process, historical data points corresponding to the abnormal distances are found, and these data points are the parts that need to be updated, generating multiple sensor data to be updated.

[0057] In the order of multiple operating cycles, the new multiple operating sensor data are mapped and replaced with the multiple sensor data to be updated. In this way, the historical data that may have problems are replaced with new accurate data, thereby updating the memory module and obtaining a memory update module. This updated module contains more accurate operating data of the hydropower unit and can better reflect the current operating status.

[0058] The memory update module is a collection of the latest and most accurate operation sensor data obtained after a series of previous processing. This module integrates historical operation sensor data and new operation sensor data, and updates and corrects data that may be abnormal, with higher reliability and representativeness. Multiple operation sensor data sets are data sets collected in real time during the operation of the current hydropower unit, including parameters measured by various sensors, such as temperature, pressure, vibration, etc.

[0059] The multiple operating sensor data sets are the data collected in real time by the sensors of the hydropower unit at the current moment. These data cover multiple key parameters such as temperature, pressure, vibration, etc., and can timely reflect the current operating status of the hydropower unit. The long short-term memory network is a deep learning model that is very suitable for processing time series data. When performing regression prediction, the data in the memory update module and multiple operating sensor data sets are input into the long short-term memory network together. The network first preprocesses the input data, such as normalization, to ensure that the data is within the appropriate range. Then, the LSTM network begins to process the input data step by step. It analyzes the trends, periodicity and other characteristics in the historical data, and combines the current operating sensor data to predict the future operating status of the hydropower unit. The memory unit plays a key role in this process. It can remember long-term dependencies, allowing the network to better understand the operating mode of the hydropower unit. As the data flows in the network, the LSTM network will continuously adjust its internal parameters to minimize the error between the predicted value and the actual value. Through repeated training and optimization, the network gradually improves the accuracy of the prediction. Finally, the LSTM network generates multiple operating status prediction results, which include specific numerical predictions of various parameters of the hydropower unit in the future, such as the temperature change trend and the pressure fluctuation range.

[0060] In a possible implementation, the operation monitoring result generating module further includes:

[0061] An operation evaluation of the hydropower unit is performed based on the operation environment data set, and an expected operation state threshold is set according to the evaluation result; it is determined whether the multiple operation state prediction results are greater than or equal to the expected operation state threshold; if the multiple operation state prediction results are greater than or equal to the expected operation state threshold, the operation state prediction results greater than or equal to the expected operation state threshold are added to the normal operation state data group, and a plurality of normal operation prediction tags are generated, and there is a corresponding relationship between the normal operation state data group and the plurality of normal operation prediction tags; if the multiple operation state prediction results are less than the expected operation state threshold, the operation state prediction results less than the expected operation state threshold are added to the abnormal operation state data group, and a plurality of abnormal operation prediction tags are generated, and there is a corresponding relationship between the abnormal operation state data group and the plurality of abnormal operation prediction tags; the plurality of normal operation prediction tags and the plurality of abnormal operation prediction tags are added to the operation monitoring results.

[0062] Specifically, a comprehensive operation evaluation of the hydropower unit is conducted using the operating environment data set, which covers various factors, such as water temperature, water flow velocity, external climate conditions, etc. Through in-depth analysis of these data and combined with the operating parameters of the hydropower unit, a comprehensive judgment is made on the operating performance of the hydropower unit under a specific environment. Based on the evaluation results, an expected operating state threshold is set. This threshold is an important criterion for measuring whether the hydropower unit is in a good operating state. Its determination requires comprehensive consideration of multiple factors to ensure its rationality and accuracy.

[0063] Judge multiple operating status prediction results, and compare each prediction result with the expected operating status threshold to see whether it is greater than or equal to this threshold. If a certain operating status prediction result meets the condition that it is greater than or equal to the expected operating status threshold, it means that the operating status of the hydropower unit corresponding to this prediction result is relatively ideal. At this time, add this prediction result to the normal operating status data group. At the same time, in order to facilitate subsequent identification and analysis, generate multiple normal operating prediction labels, and ensure that there is a corresponding relationship between the normal operating status data group and these labels. In this way, when needed, the corresponding normal operating status data can be quickly found through the label.

[0064] If multiple operation state prediction results are less than the expected operation state threshold, it indicates that there may be problems with the operation state of the hydropower unit corresponding to the prediction result. Therefore, the prediction result less than the expected operation state threshold is added to the abnormal operation state data group. Similarly, in order to facilitate the investigation and processing of abnormal operation states, multiple abnormal operation prediction labels are generated, and it is ensured that there is a corresponding relationship between the abnormal operation state data group and these labels.

[0065] Finally, multiple normal operation prediction labels and multiple abnormal operation prediction labels are added to the operation monitoring results. The operation monitoring results can clearly display the operation status classification of the hydropower unit. By viewing the operation monitoring results, the operation status of the hydropower unit can be quickly understood. The normal operation status can be continuously monitored and optimized, and the abnormal operation status can be diagnosed and repaired in time to ensure the safe, stable and efficient operation of the hydropower unit.

[0066] In a possible implementation, the state prediction optimization channel generation module further includes:

[0067] Extract the multiple abnormal operation prediction tags, traverse the multiple operation state prediction results, and determine multiple abnormal operation state operation results; perform deviation calculation based on the multiple abnormal operation state operation results to generate multiple abnormal deviation values; backtrack the multiple operation sensor data according to the multiple abnormal deviation values ​​to determine multiple abnormal data sources, compare and verify the multiple abnormal data sources with the multiple operation state prediction results, and generate multiple abnormal prediction verification results; add the multiple abnormal prediction verification results to the prediction feedback information.

[0068] Specifically, multiple abnormal operation prediction labels are extracted, and multiple operation state prediction results are traversed using these labels. During the traversal process, it is possible to accurately determine which operation state prediction results belong to abnormal operation states, thereby obtaining multiple abnormal operation state operation results. These results represent the operation states of hydropower units that may have problems, providing specific targets for subsequent analysis.

[0069] After determining multiple abnormal operating state results, deviation calculation is performed to more accurately quantify the degree of these abnormal states. It is necessary to determine a reference standard for normal operating state, based on the stable operating range in historical data, theoretical model prediction values, or pre-set specification ranges, and then compare each abnormal operating state result with the reference standard for normal operating state. By calculating the difference operation, the degree of deviation between the abnormal result and the normal state is determined, thus generating multiple abnormal deviation values, each of which represents the difference between an abnormal operating state and a normal state.

[0070] Based on multiple abnormal deviation values, multiple operating sensor data are traced back. In this process, the relationship between each abnormal deviation value and the operating sensor data is analyzed. By comparing the data changes at different time points and the data performance of each sensor, the source of these abnormal deviations is determined, so as to find multiple abnormal data sources. These abnormal data sources may be a failure of a specific sensor, interference in the data transmission process, or abnormal operation of a component inside the hydropower unit. The determined multiple abnormal data sources are compared and verified with multiple operating state prediction results, and the features and information extracted from the abnormal data sources are compared with the previously predicted operating state results to check whether the two are consistent. If they are consistent, it means that the prediction results accurately reflect the source of the abnormal situation; if they are inconsistent, the reasons need to be further analyzed, which may be that the determination of the abnormal data source is inaccurate or the prediction model has certain limitations. After comparison and verification, multiple abnormal prediction verification results are generated to help accurately locate the problem and take effective measures to solve the abnormal operating state of the hydropower unit.

[0071] Adding multiple abnormal prediction verification results to the prediction feedback information is to enrich and improve the evaluation and monitoring of the operating status of the hydropower unit. These abnormal prediction verification results include a detailed analysis of the abnormal operating status of the hydropower unit, such as the specific manifestations of the abnormality, possible abnormal data sources, and comparisons with the operating status prediction results. When these results are added to the prediction feedback information, the prediction feedback information can more comprehensively reflect the actual operating conditions of the hydropower unit. By viewing the prediction feedback information, we can quickly understand the problems existing in the hydropower unit and the severity of the problems, so as to take appropriate measures in time to troubleshoot and repair the faults, so as to ensure the safe and stable operation of the hydropower unit. At the same time, the prediction feedback information with the addition of abnormal prediction verification results can also provide valuable references for subsequent data analysis and model optimization, and help continuously improve the operation monitoring and prediction system of the hydropower unit.

[0072] In a possible implementation, the state prediction optimization channel generation module further includes:

[0073] Based on the multiple abnormal operating state operation results, cluster analysis is performed according to the multiple abnormal deviation values ​​to determine multiple deviation types; based on the multiple deviation types and the multiple abnormal deviation values, the multiple operation sensor data are traced back according to the multiple operation cycles to draw a trend chart of operation sensor data changes; the operation sensor data change trend chart is traversed to locate fluctuations and determine the multiple abnormal data sources.

[0074] Specifically, when faced with multiple abnormal operating state results, cluster analysis is performed based on multiple abnormal deviation values, and the abnormal deviation values ​​corresponding to each abnormal operating state result are carefully studied. These deviation values ​​reflect the degree of difference between the abnormal state and the normal state. Then, a clustering algorithm is used to group these abnormal results according to the similarity of the deviation values. For example, those abnormal operating states with relatively close deviation values ​​are classified into the same category, thereby determining multiple deviation types. Such classification helps to better understand different types of abnormal situations, and it will be found that some abnormalities are caused by similar reasons, such as specific fault types of sensors, performance degradation of specific components of hydropower units, etc. By determining multiple deviation types, a more targeted direction is provided for subsequent fault diagnosis and processing, and abnormal problems of hydropower units can be solved more efficiently.

[0075] According to multiple deviation types and corresponding multiple abnormal deviation values, multiple operating sensor data are traced back according to multiple operating cycles. In this process, the data collected by each sensor in each operating cycle are analyzed one by one, and the relationship between the change characteristics of the data and the deviation in a specific operating cycle is determined by combining different deviation types and abnormal deviation values. By integrating this information, a trend chart of the change of operating sensor data is drawn. This trend chart can intuitively show how the data of each sensor changes over time in different operating cycles, and the correspondence with different deviation types and abnormal deviation values. For example, for a specific deviation type, it can be clearly seen in the trend chart that the data of the relevant sensors show an upward, downward or fluctuating trend in a specific operating cycle, so as to better understand the change process of the operating status of the hydropower unit and provide important visualization basis for further analysis of the root cause of the abnormal situation.

[0076] When processing the trend graph of the change of operating sensor data, it is a key step to locate the fluctuation by traversing this trend graph. Carefully observe the changes of each data point in the trend graph, and gradually analyze the trend of the data along the time axis. In this process, pay special attention to those places where the data fluctuates significantly. These fluctuations are manifested as a sharp rise, fall or irregular change pattern of the data. For each possible fluctuation point, compare the data performance of different sensors at the same time point or time period. If the data of multiple sensors fluctuate at the same time, then this area is more likely to be the source of the abnormality. At the same time, combined with the previously determined deviation type and abnormal deviation value, further analyze the relationship between these fluctuations and abnormal conditions. For example, if the data of the temperature sensor suddenly rises in a certain time period, and the data of the pressure sensor also changes abnormally, and the deviation type and abnormal deviation value corresponding to this time period also show a large deviation, then it can be preliminarily determined that this time period and the related sensors are abnormal data sources. Through a comprehensive traversal and analysis of the entire trend graph of the change of operating sensor data, multiple abnormal data sources are determined. These abnormal data sources are a failure of a specific component in the hydropower unit, such as abnormal fluctuations in the data of the temperature and vibration sensors caused by bearing wear; or they may be affected by external environmental factors, such as changes in the data of the pressure sensor caused by changes in water flow. After these abnormal data sources are determined, a clear direction is provided for subsequent fault diagnosis and repair work, and more targeted measures can be taken to improve the operating stability and reliability of the hydropower units.

[0077] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description and does not represent the advantages and disadvantages of the embodiments. And the above-mentioned specific embodiments of this specification are described. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0078] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application should be included in the protection scope of the present application.

[0079] This specification and the drawings are merely exemplary illustrations of the present application and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. A hydropower unit state prediction system based on deep learning, characterized in that: The system comprises: An operation sensor data acquisition module, wherein the operation sensor data acquisition module collects data of the hydropower unit in real time through a data sensor network based on multiple operation periods to obtain multiple operation sensor data; A state prediction channel construction module, the state prediction channel construction module is used to perform deep learning according to the historical operation sensor data set according to the multiple operation time periods to construct a state prediction channel; An operation monitoring result generating module, the operation monitoring result generating module is used to synchronize the multiple operation sensor data to the state prediction channel for analysis, generate multiple operation state prediction results, perform operation monitoring on the hydropower unit according to the multiple operation state prediction results combined with the operation environment data set, and generate an operation monitoring result, wherein the operation monitoring result includes multiple state tags; A state prediction optimization channel generation module, the state prediction optimization channel generation module is used to traverse the multiple running state prediction results according to the multiple state tags to verify, generate prediction feedback information, update and optimize the state prediction channel according to the prediction feedback information, and generate a state prediction optimization channel; An intelligent prediction module is used to perform intelligent prediction of the state of the hydropower unit through the state prediction optimization channel.

2. The hydropower unit state prediction system based on deep learning according to claim 1, characterized in that: The state prediction channel building module also includes: Performing operation analysis on the hydropower unit based on the multiple operation time periods, and generating multiple operation cycles according to the analysis results; Retrieving a historical operation data archive of the hydropower unit, extracting the historical operation sensor data set according to the historical operation data archive, performing feature extraction on the historical operation sensor data set according to the multiple operation cycles, and generating multiple historical operation features; Performing time series analysis according to the multiple operation cycles combined with the multiple historical operation characteristics to determine the operation time step of the hydropower unit; Using a long short-term memory network to perform deep learning on the historical operation sensor data set according to the operation time step to generate an operation time series, wherein the operation time series includes a plurality of operation time nodes, and each of the plurality of operation time nodes has at least one historical operation sensor data; The plurality of historical operation sensor data are verified and evaluated according to the plurality of operation time nodes, a verification token is generated, an evaluation is performed according to the verification token, and a state prediction channel is output.

3. The hydropower unit state prediction system based on deep learning according to claim 2, characterized in that: The state prediction channel building module also includes: Performing data time series normalization processing on the historical operation sensor data set according to the operation time step to generate a historical sensor standard data set; Segmenting the historical sensing standard data set according to the running time step to generate a plurality of standard data sample segments; Performing a long-term analysis based on the multiple standard data sample segments to generate a first analysis result, extracting multiple long-term standard data sample segments according to the first analysis result, and performing calculations based on the multiple long-term standard data sample segments to generate a long-term dependence coefficient; Performing short-term analysis based on the multiple standard data sample segments to generate a second analysis result, extracting multiple short-term standard data sample segments according to the second analysis result, and performing calculations based on the multiple short-term standard data sample segments to generate a short-term dependence coefficient; Capturing the historical operation sensor data set according to the long-term dependency coefficient to generate a first information flow; capturing the historical operation sensor data set according to the short-term dependency coefficient to generate a second information flow; The first information flow and the second information flow are stacked to determine a plurality of running time nodes, and the running time sequence is extracted by integrating the plurality of running time nodes.

4. The hydropower unit state prediction system based on deep learning according to claim 3, characterized in that: The state prediction channel building module also includes: Aligning the first information flow with the second information flow according to the running time step to generate a timestamp alignment result; Based on the timestamp alignment result, stacking the first information flow and the second information flow according to the time dimension to generate a two-dimensional data matrix; Traversing the two-dimensional data matrix to perform information change analysis, generate an information change curve, and extract the multiple running time nodes according to the information change curve; A multi-dimensional integration is performed according to the multiple running time nodes to generate a multi-dimensional vector set, and the multi-dimensional vector set is arranged according to the multiple running cycles to determine the running time sequence.

5. The hydropower unit state prediction system based on deep learning according to claim 2, characterized in that: The state prediction channel building module also includes: Synchronizing the plurality of operation sensor data to the state prediction channel and storing them in a memory module of the state prediction channel, wherein the memory module includes the historical operation sensor data set; Performing distance calculation on the plurality of operation sensor data and the historical operation sensor data set through the memory module to determine a plurality of Euclidean distance data; Determine whether the plurality of Euclidean distance data are in an expected distance interval, extract the Euclidean distance data that are not in the expected distance interval, and determine a plurality of abnormal distances; Using the multiple abnormal distances as indexes, traversing the historical operation sensor data set to generate multiple sensor data to be updated; According to the multiple operation cycles, mapping and replacing the multiple operation sensor data with the multiple sensor data to be updated, and updating the memory module to obtain a memory update module; Regression prediction is performed based on the memory update module and the multiple operation sensor data sets to generate the multiple operation state prediction results.

6. The hydropower unit state prediction system based on deep learning according to claim 2, characterized in that: The operation monitoring result generating module also includes: Performing an operation evaluation on the hydropower unit based on the operation environment data set, and setting an expected operation state threshold according to the evaluation result; Determine whether the multiple running state prediction results are greater than or equal to the expected running state threshold; If the multiple operating state prediction results are greater than or equal to the expected operating state threshold, the operating state prediction results greater than or equal to the expected operating state threshold are added to the normal operating state data group, and a plurality of normal operating prediction labels are generated, and there is a corresponding relationship between the normal operating state data group and the plurality of normal operating prediction labels; If the multiple running state prediction results are less than the expected running state threshold, the running state prediction results less than the expected running state threshold are added to the abnormal running state data group, and a plurality of abnormal running prediction labels are generated, and there is a corresponding relationship between the abnormal running state data group and the plurality of abnormal running prediction labels; The multiple normal operation prediction labels and the multiple abnormal operation prediction labels are added to the operation monitoring result.

7. The hydropower unit state prediction system based on deep learning according to claim 6, characterized in that: The state prediction optimization channel generation module also includes: Extract the multiple abnormal operation prediction tags, traverse the multiple operation state prediction results, and determine multiple abnormal operation state operation results; Perform deviation calculation based on the multiple abnormal operation state operation results to generate multiple abnormal deviation values; Backtracking the multiple operating sensor data according to the multiple abnormal deviation values ​​to determine multiple abnormal data sources, and comparing and verifying the multiple abnormal data sources with the multiple operating state prediction results to generate multiple abnormal prediction verification results; The multiple abnormal prediction verification results are added to the prediction feedback information.

8. The hydropower unit state prediction system based on deep learning according to claim 7, characterized in that: The state prediction optimization channel generation module also includes: Performing cluster analysis based on the multiple abnormal operation state operation results according to the multiple abnormal deviation values ​​to determine multiple deviation types; Backtracking the plurality of operation sensor data according to the plurality of operation cycles according to the plurality of deviation types and the plurality of abnormal deviation values, and drawing a change trend graph of the operation sensor data; The operation sensor data change trend graph is traversed to locate fluctuations and determine the multiple abnormal data sources.

Images