Hydropower station maintenance method and system based on multiple parameters

Abstract

The invention discloses a hydropower station maintenance method and system based on multiple parameters, and relates to the technical field of hydropower station maintenance. The invention provides a maintenance method and system by collecting meteorological data, geological data and monitoring data and analyzing the influence of rockfall on the Jiwu hydropower station from multiple dimensions, and solves the problem that there is no maintenance method for the Jiwu hydropower station by rockfall at present, the risk of damage to hydropower station equipment is effectively reduced, the equipment of the Jiwu hydropower station is prolonged, and the service life of the Jiwu hydropower station is prolonged. The service life of the water turbine is prolonged; according to the method, the corresponding maintenance scheme is formulated according to statistics of the number and the size of the rockfall, the operation efficiency of hydropower station equipment can be improved, the time of power failure caused by the rockfall and other problems is shortened, the cost of manual intervention is greatly reduced, and the safety of maintenance work is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of hydropower station maintenance, and in particular to a multi-parameter-based hydropower station maintenance method and system. Background Art

[0002] The Jiniu Hydropower Station is located in a valley in a certain area of ​​my country. The terrain here is complex, with steep mountains, rocks and soil. This terrain feature makes the operation of the hydropower station face huge challenges. Especially in the rainy season or when natural disasters such as earthquakes occur frequently, it is easy to cause disasters such as landslides and mud-rock flows, which will cause serious damage to the equipment and facilities of the hydropower station. Among them, the water inlet and water inlet pipeline are among the most vulnerable facilities. Due to the blockage of the water inlet and the breakage of the water inlet pipeline, the hydropower station may be cut off from power, seriously affecting the local electricity demand.

[0003] However, existing maintenance methods mainly rely on manual inspections and regular cleaning, which is not only inefficient but also prone to errors. Especially during the rainy season and when natural disasters occur, manual inspections and cleaning are often not completed in time, which poses a huge risk to the safe operation of hydropower stations.

[0004] At present, there is no maintenance method for Jiniu Hydropower Station caused by falling rocks in my country, which has increased the risk of damage to hydropower station equipment to a certain extent. Therefore, it is particularly important to research and develop a technology that can automatically identify and remove falling rocks. Such a technology can not only improve the operating efficiency of hydropower station equipment and reduce the time of power outages caused by falling rocks and other problems, but also greatly reduce the cost of manual intervention and improve the safety of maintenance work. Summary of the invention

[0005] The object of the present invention is to provide a multi-parameter based hydropower station maintenance method and system to improve the above technical problems.

[0006] In order to achieve the above-mentioned object of the invention, the embodiment of the present invention provides the following technical solutions:

[0007] A multi-parameter-based hydropower station maintenance method, comprising:

[0008] Collecting actual operation data of a hydropower station; calculating actual power generation efficiency of the hydropower station based on the actual operation data;

[0009] Acquire historical rockfall event data, monitoring data, geological data and meteorological data of the hydropower station and pre-process the historical rockfall event data, the geological data and the meteorological data;

[0010] Determining a preset power generation threshold of the hydropower station; and judging a current state of the hydropower station based on the preset power generation threshold and the actual power generation efficiency;

[0011] Constructing a maintenance model; wherein the maintenance model includes a rockfall judgment model and a rockfall disaster prediction model;

[0012] Inputting the pre-processed historical rockfall event data into the rockfall disaster prediction model for training to obtain a trained rockfall disaster prediction model;

[0013] According to the current state of the hydropower station, the monitoring data is input into the rockfall judgment model to obtain a corresponding judgment result;

[0014] According to the current state of the hydropower station, the pre-processed geological data and meteorological data are input into the trained rockfall disaster prediction model to obtain corresponding prediction results;

[0015] A maintenance plan for the hydropower station is formulated based on the judgment result and the prediction result.

[0016] Furthermore, the actual operation data include water flow velocity, water flow cross-sectional area, and working water head data; the historical rockfall event data include historical geological data and historical meteorological data of rockfall disasters; the historical geological data and the geological data both include rock and soil mechanical parameters, geological structure data, and soil conditions; the meteorological data are weather data for the next month; the historical meteorological data and the weather data both include daily rainfall, cumulative rainfall, rainfall intensity, rainfall duration, and temperature data; the monitoring data are monitoring video data at the water inlet and turbine of the hydropower station.

[0017] Further, the calculating the actual power generation efficiency of the hydropower station based on the actual operation data includes:

[0018] Based on the water flow velocity and the water flow cross-sectional area, the corresponding power generation flow of the hydropower station is calculated; based on the power generation flow and the working water head data, the corresponding theoretical power generation efficiency of the hydropower station is calculated;

[0019] The operating environment of the hydropower station and the corresponding turbine efficiency and generator efficiency are collected to determine the corresponding power generation efficiency coefficient; based on the power generation efficiency coefficient and the theoretical power generation efficiency, the actual power generation efficiency of the hydropower station is calculated.

[0020] Furthermore, the preprocessing includes data cleaning, feature selection and normalization processing: wherein:

[0021] Data cleaning is used to detect whether the historical rockfall event data, geological data and meteorological data have missing values, duplicate data and outliers; if there are missing values, fill them with interpolation methods; if there are duplicate data and outliers, remove the corresponding data;

[0022] Feature selection is used to process the historical rockfall event data, geological data and meteorological data after data cleaning using the cross-validation recursive feature elimination method to obtain the corresponding feature data;

[0023] Normalization processing is used to scale each of the characteristic data to a range of 0 to 1 to obtain pre-processed historical rockfall event data, geological data and meteorological data.

[0024] Further, the determining of a preset power generation threshold of the hydropower station; and judging a current state of the hydropower station based on the preset power generation threshold and the actual power generation efficiency include:

[0025] Collect the historical actual power generation efficiency of the hydropower station every day within one year; draw a box plot corresponding to each of the historical actual power generation efficiencies; use the interquartile range to set a threshold corresponding to the historical actual power generation efficiency, remove the historical actual power generation efficiency exceeding the threshold as an outlier, and obtain the screened historical actual power generation efficiency; calculate the average value of the screened historical actual power generation efficiency and use it as the preset power generation threshold of the hydropower station;

[0026] Determine whether the actual power generation efficiency of the hydropower station is less than the preset power generation threshold of the hydropower station; if so, the current state of the hydropower station is "fault detection required" and enters the next operation; otherwise, the current state of the hydropower station is "normal" and the original maintenance plan is executed.

[0027] Furthermore, the rockfall judgment model includes an image extraction module, a target detection module, a rockfall statistics module, and a rockfall judgment module connected in series in sequence; the target detection module adopts a YOLO8v model; the rockfall judgment module adopts a CNN model; the rockfall disaster prediction model adopts a long short-term memory network;

[0028] The YOLOv8 network model includes a backbone network Backbone, a neck network Neck and a head network Head; the backbone network Backbone includes a CBS layer, a first module, a second module, a third module, a fourth module and an SPPF layer connected in series in sequence; the first module, the second module and the third module are all CBS-C2f modules, which include a CBS layer and a C2f layer connected in series; the fourth module is a CBS-SEMP-C2f module, which includes a CBS layer and a SEMP-C2f submodule connected in series; the SEMP-C2f submodule includes a Conv layer, a split layer, two SEMP-Bottleneck layers, a Concat layer and a Conv layer connected in series in sequence; the SEMP-Bottleneck layer The ottleneck layer includes a series of Conv layers, a mixed attention layer and a Conv layer, and the mixed attention layer includes a parallel SE attention layer and an MPCA attention layer; the neck network Neck includes a fifth module and a sixth module; the fifth module includes an Upsample layer, a C2f layer and an Upsample layer connected in series in sequence; the sixth module includes a seventh module, an eighth module and a C2f layer connected in series; the seventh module and the eighth module both include a C2f layer and a CBS layer connected in series; the head network Head includes three parallel Detect layers; wherein the SE attention layer adopts a Squeeze-and-Excitation channel attention mechanism, and the MPCA attention layer adopts a multi-scale position attention mechanism.

[0029] Furthermore, the rockfall disaster prediction model is trained, including:

[0030] The preprocessed historical geological data and historical meteorological data are input into the long short-term memory network, and the initial model parameters of the long short-term memory network are determined by a cross-validation method; the initial model parameters are applied to the long short-term memory network, and the parameters of the long short-term memory network are adjusted using a grid search method to obtain a trained rockfall disaster prediction model.

[0031] Furthermore, the method for obtaining the judgment result includes:

[0032] Input the monitoring data into the image extraction module for processing; divide the monitoring video data into video data with a time length of 1 hour; set the frame rate, and obtain the number of monitoring image data corresponding to the monitoring video data according to the frame rate and the time length; perform image division on the monitoring video data according to the number of monitoring image data to obtain corresponding monitoring image data;

[0033] Input the monitoring image data into the YOLO model for processing to obtain multiple initial detection frames; calculate the confidence of the multiple initial detection frames, and respectively determine whether the confidence corresponding to the multiple initial detection frames exceeds a preset confidence threshold; if so, retain the corresponding initial detection frame; otherwise, discard the corresponding initial detection frame to obtain a final detection frame;

[0034] The final detection frame and the monitoring image data are input into the CNN model to determine the corresponding rockfall detection result; according to the rockfall detection result, each final detection frame is traversed and the centroid coordinates of the corresponding internal pixel values ​​are calculated; based on the intersection of each centroid coordinate and the corresponding final detection frame boundary, the spatial position of each rockfall is calculated; according to the spatial position, the number of rocks in each final detection frame is calculated to obtain the total number of rocksfall; according to the total number of rocksfall and the pixel value of the corresponding monitoring image data, the total volume of rocksfall is calculated;

[0035] Based on the water flow velocity of the hydropower station, the total volume of the falling rocks is processed by the falling rocks judgment module; when the water flow velocity of the hydropower station is greater than a preset water flow threshold, if so, it is determined that there is a risk to the turbine of the hydropower station, and the risk level is determined according to the total volume of the falling rocks; otherwise, it is determined whether the total volume of the falling rocks reaches the preset volume threshold, if so, it is determined that the water inlet of the hydropower station needs to be cleaned, otherwise, it is determined that the water inlet of the hydropower station does not need to be cleaned; wherein, the risk levels include high, medium and low; the judgment result includes the risk level and whether the water inlet needs to be cleaned.

[0036] Furthermore, formulating a maintenance plan for the hydropower station according to the judgment result and the prediction result includes:

[0037] Determine whether any two of the following conditions are met: the prediction result exceeds the first threshold, the risk level is medium or high, and the water inlet of the hydropower station needs to be cleaned; if so, quickly arrange technicians to clean the water inlet, check the status of the turbine of the hydropower station and perform repairs, and check and strengthen the stability of the surrounding protective net;

[0038] Determine whether the prediction result exceeds the second threshold value and satisfies any one of the following conditions: the risk level is medium or high, and the water inlet of the hydropower station needs to be cleaned; then arrange technicians to check the stability of the surrounding protective nets and check or repair the status of the turbine or clean the water inlet;

[0039] In other cases, the original maintenance plan is implemented.

[0040] A multi-parameter-based hydropower station maintenance system includes a data acquisition module, a data processing module, a state determination module, a rockfall statistics module, a rockfall disaster prediction module, and a maintenance plan formulation module; the data processing module includes a data preprocessing submodule and a threshold determination submodule; wherein:

[0041] A data acquisition module is used to obtain the actual operation data of the hydropower station, historical rockfall event data, monitoring data, geological data and meteorological data;

[0042] A data preprocessing submodule, for performing data cleaning, feature selection and normalization processing on the historical rockfall event data, the geological data and the meteorological data;

[0043] A threshold determination submodule, used to calculate the actual power generation efficiency of the hydropower station based on the actual operation data; and determine a preset power generation threshold of the hydropower station;

[0044] A state determination module, used for determining the current state of the hydropower station based on the preset power generation threshold and the actual power generation efficiency;

[0045] A rockfall statistics module is used to process the monitoring data through the rockfall judgment model according to the current state of the hydropower station to obtain a corresponding judgment result;

[0046] A rockfall disaster prediction module is used to process the preprocessed geological data and the preprocessed meteorological data using the trained rockfall disaster prediction model according to the current state of the hydropower station to obtain corresponding prediction results;

[0047] The maintenance plan formulation module is used to formulate a maintenance plan for the hydropower station based on the judgment result and the prediction result.

[0048] The present invention collects meteorological data, geological data and monitoring data, analyzes the impact of rockfall on the Jiniu Hydropower Station from multiple dimensions, and proposes a maintenance method and system, which solves the problem that there is currently no maintenance method for the Jiniu Hydropower Station caused by rockfall, effectively reduces the risk of damage to hydropower station equipment, and extends the service life of equipment at the Jiniu Hydropower Station, such as turbines; the present invention formulates corresponding maintenance plans based on statistics on the number and volume of rockfalls, which can improve the operating efficiency of hydropower station equipment, reduce the time of power outages caused by rockfalls and other problems, greatly reduce the cost of manual intervention, and improve the safety of maintenance work. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0050] Figure 1 is a flow chart of the method of the present invention;

[0051] Figure 2 It is a structural schematic diagram of the target detection module of the present invention;

[0052] Figure 3 It is a schematic diagram of the structure of CBS-C2f and CBS-SEMP-C2f of the present invention;

[0053] Figure 4 It is a schematic diagram of the system structure of the present invention. DETAILED DESCRIPTION

[0054] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents the selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present invention.

[0055] See also Figure 1 The present embodiment provides a multi-parameter-based hydropower station maintenance method, which is characterized by comprising:

[0056] Collecting actual operation data of a hydropower station; calculating actual power generation efficiency of the hydropower station based on the actual operation data;

[0057] The actual operation data include water flow velocity, water flow cross-sectional area, and working water head data;

[0058] Based on the water flow velocity and the water flow cross-sectional area, the corresponding power generation flow of the hydropower station is calculated; based on the power generation flow and the working water head data, the corresponding theoretical power generation efficiency of the hydropower station is calculated;

[0059] The operating environment of the hydropower station and the corresponding turbine efficiency and generator efficiency are collected to determine the corresponding power generation efficiency coefficient; based on the power generation efficiency coefficient and the theoretical power generation efficiency, the actual power generation efficiency of the hydropower station is calculated.

[0060] Acquire historical rockfall event data, monitoring data, geological data and meteorological data of the hydropower station and pre-process the historical rockfall event data, the geological data and the meteorological data;

[0061] The historical rockfall event data include historical geological data and historical meteorological data of rockfall disasters; both the historical geological data and the geological data include rock and soil mechanical parameters, geological structure data, and soil conditions; the meteorological data are weather data for the next month; both the historical meteorological data and the weather data include daily rainfall, cumulative rainfall, rainfall intensity, rainfall duration, and temperature data; the monitoring data are monitoring video data at the water inlet and turbine of the hydropower station.

[0062] The preprocessing includes data cleaning, feature selection and normalization processing:

[0063] Data cleaning is used to detect whether the historical rockfall event data, geological data and meteorological data have missing values, duplicate data and outliers; if there are missing values, fill them with interpolation methods; if there are duplicate data and outliers, remove the corresponding data;

[0064] Feature selection is used to process the historical rockfall event data, geological data and meteorological data after data cleaning using the cross-validation recursive feature elimination method to obtain the corresponding feature data;

[0065] Normalization processing is used to scale each of the characteristic data to a range of 0 to 1 to obtain pre-processed historical rockfall event data, geological data and meteorological data.

[0066] Determining a preset power generation threshold of the hydropower station; and judging a current state of the hydropower station based on the preset power generation threshold and the actual power generation efficiency, which includes:

[0067] Collect the historical actual power generation efficiency of the hydropower station every day within one year; use the historical actual power generation efficiency as data points to draw box plots corresponding to each historical actual power generation efficiency; use the interquartile range to formulate the corresponding threshold, that is, obtain the minimum value, the first quartile Q1, the median Q2, the third quartile Q3 and the maximum value of all data points, use the difference between the third quartile Q3 and the first quartile Q1 as the interquartile range IQR, set Q1-1.5IQR and Q3+1.5IQR as thresholds, and remove data points (historical actual power generation efficiency) outside Q1-1.5IQR and Q3+1.5IQR as outliers to obtain the screened historical actual power generation efficiency; calculate the average value of the screened historical actual power generation efficiency and use it as the preset power generation threshold of the hydropower station;

[0068] Determine whether the actual power generation efficiency of the hydropower station is less than the preset power generation threshold of the hydropower station; if so, the current state of the hydropower station is "need to perform fault detection" and enter the next operation; otherwise, the current state of the hydropower station is "normal" and the original maintenance plan is executed. The original maintenance plan refers to the initial maintenance plan formulated according to the working state of the initial Jiniu Hydropower Station or the maintenance plan formulated by the last maintenance of the Jiniu Hydropower Station using the present invention.

[0069] Constructing a maintenance model; wherein the maintenance model includes a rockfall judgment model and a rockfall disaster prediction model;

[0070] The rockfall judgment model includes an image extraction module, a target detection module, a rockfall statistics module, and a rockfall judgment module connected in series in sequence; the target detection module adopts a YOLO model; the rockfall judgment module adopts a CNN model; and the rockfall disaster prediction model adopts a long short-term memory network.

[0071] Inputting the pre-processed historical rockfall event data into the rockfall disaster prediction model for training to obtain a trained rockfall disaster prediction model;

[0072] The rockfall disaster prediction model is trained, including:

[0073] The pre-processed historical geological data and historical meteorological data are input into the long short-term memory network, and the initial model parameters of the long short-term memory network are determined by a cross-validation method. The cross-validation method can prevent the model from overfitting and improve the generalization ability of the long short-term memory network; the initial model parameters are applied to the long short-term memory network, and the grid search method is used to adjust the parameters of the long short-term memory network, adjust the learning rate of the long short-term memory network, the number of neurons in the hidden layer, the number of layers, the batch size and other parameters, and select the best performance parameter combination according to the results of the grid search, that is, obtain the trained rockfall disaster prediction model. After each training, the mean square error can be selected to measure the prediction accuracy of the rockfall disaster prediction model.

[0074] According to the current state of the hydropower station, the monitoring data is input into the rockfall judgment model to obtain a corresponding judgment result;

[0075] The method for obtaining the judgment result includes:

[0076] Input the monitoring data into the image extraction module for processing; divide the monitoring video data into video data with a time length of 1 hour; set the frame rate, and obtain the number of monitoring image data corresponding to the monitoring video data according to the frame rate and the time length; perform image division on the monitoring video data according to the number of monitoring image data to obtain the corresponding monitoring image data; wherein, before data division, the monitoring video data can be subjected to image enhancement, image denoising and other processing, which can improve the efficiency and accuracy of feature extraction of subsequent models to a certain extent, and help detect rockfalls and calculate the number and volume of rockfalls. The setting of the time length can also be set according to actual needs.

[0077] Input the monitoring image data into the YOLO model for processing to obtain multiple initial detection frames; calculate the confidence of the multiple initial detection frames, and respectively determine whether the confidence corresponding to the multiple initial detection frames exceeds a preset confidence threshold; if so, retain the corresponding initial detection frame; otherwise, discard the corresponding initial detection frame to obtain a final detection frame;

[0078] like Figure 3 and Figure 4As shown, the YOLO model is a YOLOv8 network model; the YOLOv8 network model includes a backbone network Backbone, a neck network Neck and a head network Head; the backbone network Backbone includes a CBS layer, a first module, a second module, a third module, a fourth module and an SPPF layer connected in series in sequence; the first module, the second module and the third module are all CBS-C2f modules, which include a CBS layer and a C2f layer connected in series; the fourth module is a CBS-SEMP-C2f module, which includes a CBS layer and a SEMP-C2f submodule connected in series; the SEMP-C2f submodule includes a Conv layer, a split layer, two SEMP-Bottleneck layers, a Concat layer and a C2f layer connected in series in sequence onv layer; the SEMP-Bottleneck layer includes a series of Conv layers, mixed attention layers and Conv layers, and the mixed attention layer includes a parallel SE attention layer and MPCA attention layer; the neck network Neck includes a fifth module and a sixth module; the fifth module includes a series of Upsample layers, C2f layers and Upsample layers; the sixth module includes a series of seventh modules, eighth modules and C2f layers; the seventh module and the eighth module both include a series of C2f layers and CBS layers; the head network Head includes three parallel Detect layers; wherein the SE attention layer adopts the Squeeze-and-Excitation channel attention mechanism, and the MPCA attention layer adopts the multi-scale position attention mechanism.

[0079] The final detection frame and the monitoring image data are input into the CNN model to determine the corresponding rockfall detection result; according to the rockfall detection result, each final detection frame is traversed and the centroid coordinates of the corresponding internal pixel values ​​are calculated; based on the intersection of each centroid coordinate and the corresponding final detection frame boundary, the spatial position of each rockfall is calculated; according to the spatial position, the number of rocks in each final detection frame is calculated to obtain the total number of rocksfall; according to the total number of rocksfall and the pixel value of the corresponding monitoring image data, the total volume of rocksfall is calculated;

[0080] Based on the water flow velocity of the hydropower station, the total volume of the falling rocks is processed by the falling rocks judgment module; when the water flow velocity of the hydropower station is greater than a preset water flow threshold, if so, it is determined that there is a risk to the turbine of the hydropower station, and the risk level is determined according to the total volume of the falling rocks; otherwise, it is determined whether the total volume of the falling rocks reaches the preset volume threshold, if so, it is determined that the water inlet of the hydropower station needs to be cleaned, otherwise, it is determined that the water inlet of the hydropower station does not need to be cleaned; wherein, the risk levels include high, medium and low; the judgment result includes the risk level and whether the water inlet needs to be cleaned.

[0081] According to the current state of the hydropower station, the pre-processed geological data and meteorological data are input into the trained rockfall disaster prediction model to obtain corresponding prediction results;

[0082] A maintenance plan for the hydropower station is formulated based on the judgment result and the prediction result.

[0083] The method of formulating a maintenance plan for the hydropower station according to the judgment result and the prediction result includes:

[0084] Determine whether any two of the following conditions are met: the prediction result exceeds the first threshold, the risk level is intermediate or advanced, and the water inlet of the hydropower station needs to be cleaned; if so, quickly arrange for technical personnel to clean the water inlet, and install a filter at the water inlet to prevent damage caused by debris entering the turbine, and then check or repair the status of the turbine of the hydropower station. If the degree of damage is higher than the set threshold, directly replace the equipment to reduce maintenance time, reduce the time of power outages caused by falling rocks and other problems, and greatly reduce the cost of manual intervention; check and strengthen the stability of the surrounding protective net or replace it according to the degree of damage of the on-site protective net, and clean up the existing fallen rocks in the surrounding environment to prevent further damage to the Jiniu Hydropower Station by falling rocks; arrange staff to conduct inspections within the next week so that the hydropower station can be maintained in time.

[0085] Determine whether the prediction result exceeds the second threshold value and satisfies any one of the following conditions: the risk level is medium or high, and the water inlet of the hydropower station needs to be cleaned; then arrange technicians to check the stability of the surrounding protective nets and check or repair the status of the turbine or clean the water inlet;

[0086] In other cases, the original maintenance plan is implemented.

[0087] See also Figure 4 The present embodiment provides a multi-parameter-based hydropower station maintenance system, which includes a data acquisition module, a data processing module, a state determination module, a rockfall statistics module, a rockfall disaster prediction module, and a maintenance plan formulation module; the data processing module includes a data preprocessing submodule and a threshold determination submodule; wherein:

[0088] A data acquisition module is used to obtain the actual operation data of the hydropower station, historical rockfall event data, monitoring data, geological data and meteorological data;

[0089] A data preprocessing submodule, for performing data cleaning, feature selection and normalization processing on the historical rockfall event data, the geological data and the meteorological data;

[0090] A threshold determination submodule, used to calculate the actual power generation efficiency of the hydropower station based on the actual operation data; and determine a preset power generation threshold of the hydropower station;

[0091] A state determination module, used for determining the current state of the hydropower station based on the preset power generation threshold and the actual power generation efficiency;

[0092] A rockfall statistics module is used to process the monitoring data through the rockfall judgment model according to the current state of the hydropower station to obtain a corresponding judgment result;

[0093] A rockfall disaster prediction module is used to process the preprocessed geological data and the preprocessed meteorological data using the trained rockfall disaster prediction model according to the current state of the hydropower station to obtain corresponding prediction results;

[0094] The maintenance plan formulation module is used to formulate a maintenance plan for the hydropower station based on the judgment result and the prediction result.

[0095] In addition, a multi-parameter-based hydropower station maintenance system also includes an early warning module. When the prediction result is high, the system will send the information to the management personnel through the network so that the management personnel can receive the information and arrange staff for protection in the first time.

[0096] The present invention collects meteorological data, geological data and monitoring data, analyzes the impact of rockfall on the Jiniu Hydropower Station from multiple dimensions, and proposes a maintenance method and system, which solves the problem that there is currently no maintenance method for the Jiniu Hydropower Station caused by rockfall, effectively reduces the risk of damage to hydropower station equipment, and extends the service life of equipment at the Jiniu Hydropower Station, such as turbines; the present invention formulates corresponding maintenance plans based on statistics on the number and volume of rockfalls, which can improve the operating efficiency of hydropower station equipment, reduce the time of power outages caused by rockfalls and other problems, greatly reduce the cost of manual intervention, and improve the safety of maintenance work.

[0097] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A multi-parameter based hydropower station maintenance method, characterized in that: include: Collect actual operation data of a hydropower station; Calculating the actual power generation efficiency of the hydropower station based on the actual operation data; Acquire historical rockfall event data, monitoring data, geological data and meteorological data of the hydropower station and pre-process the historical rockfall event data, the geological data and the meteorological data; Determining a preset power generation threshold of the hydropower station; and judging a current state of the hydropower station based on the preset power generation threshold and the actual power generation efficiency; Constructing a maintenance model; wherein the maintenance model includes a rockfall judgment model and a rockfall disaster prediction model; Inputting the pre-processed historical rockfall event data into the rockfall disaster prediction model for training to obtain a trained rockfall disaster prediction model; According to the current state of the hydropower station, the monitoring data is input into the rockfall judgment model to obtain a corresponding judgment result; According to the current state of the hydropower station, the pre-processed geological data and meteorological data are input into the trained rockfall disaster prediction model to obtain corresponding prediction results; A maintenance plan for the hydropower station is formulated based on the judgment result and the prediction result.

2. A multi-parameter based hydropower station maintenance method according to claim 1, characterized in that: The actual operation data include water flow velocity, water flow cross-sectional area, and working water head data; the historical rockfall event data include historical geological data and historical meteorological data of rockfall disasters; the historical geological data and the geological data both include rock and soil mechanical parameters, geological structure data, and soil conditions; the meteorological data are weather data for the next month; the historical meteorological data and the weather data both include daily rainfall, cumulative rainfall, rainfall intensity, rainfall duration, and temperature data; the monitoring data are monitoring video data at the water inlet and turbine of the hydropower station.

3. A multi-parameter based hydropower station maintenance method according to claim 2, characterized in that: The calculating the actual power generation efficiency of the hydropower station based on the actual operation data comprises: Based on the water flow velocity and the water flow cross-sectional area, the corresponding power generation flow of the hydropower station is calculated; based on the power generation flow and the working water head data, the corresponding theoretical power generation efficiency of the hydropower station is calculated; The operating environment of the hydropower station and the corresponding turbine efficiency and generator efficiency are collected to determine the corresponding power generation efficiency coefficient; based on the power generation efficiency coefficient and the theoretical power generation efficiency, the actual power generation efficiency of the hydropower station is calculated.

4. A multi-parameter based hydropower station maintenance method according to claim 2, characterized in that: The preprocessing includes data cleaning, feature selection and normalization processing: Data cleaning is used to detect whether the historical rockfall event data, geological data and meteorological data have missing values, duplicate data and outliers; if there are missing values, fill them with interpolation methods; if there are duplicate data and outliers, remove the corresponding data; Feature selection is used to process the historical rockfall event data, geological data and meteorological data after data cleaning using the cross-validation recursive feature elimination method to obtain the corresponding feature data; Normalization processing is used to scale each of the characteristic data to a range of 0 to 1 to obtain pre-processed historical rockfall event data, geological data and meteorological data.

5. A multi-parameter based hydropower station maintenance method according to claim 2, characterized in that: Determining a preset power generation threshold of the hydropower station; Based on the preset power generation threshold and the actual power generation efficiency, determining the current state of the hydropower station includes: Collect the historical actual power generation efficiency of the hydropower station every day within one year; use each of the historical actual power generation efficiencies as a data point and draw a box plot corresponding to each of the historical actual power generation efficiencies; use the interquartile range to set the corresponding threshold, that is, obtain the first quartile, median, third quartile and maximum value in the box plot, and the quarterback removes the historical actual power generation efficiency exceeding the threshold as an outlier to obtain the screened historical actual power generation efficiency; calculate the average value of the screened historical actual power generation efficiency and use it as the preset power generation threshold of the hydropower station; Determine whether the actual power generation efficiency of the hydropower station is less than the preset power generation threshold of the hydropower station; if so, the current state of the hydropower station is "fault detection required" and enters the next operation; otherwise, the current state of the hydropower station is "normal" and the original maintenance plan is executed.

6. A multi-parameter based hydropower station maintenance method according to claim 2, characterized in that: The rockfall judgment model includes an image extraction module, a target detection module, a rockfall statistics module, and a rockfall judgment module connected in series; the target detection module adopts a YOLO8v network model; the rockfall judgment module adopts a CNN model; the rockfall disaster prediction model adopts a long short-term memory network; The YOLOv8 network model includes a backbone network Backbone, a neck network Neck and a head network Head; the backbone network Backbone includes a CBS layer, a first module, a second module, a third module, a fourth module and an SPPF layer connected in series in sequence; the first module, the second module and the third module are all CBS-C2f modules, which include a CBS layer and a C2f layer connected in series; the fourth module is a CBS-SEMP-C2f module, which includes a CBS layer and a SEMP-C2f submodule connected in series; the SEMP-C2f submodule includes a Conv layer, a split layer, two SEMP-Bottleneck layers, a Concat layer and a Conv layer connected in series in sequence; the SEMP-Bottleneck layer The ottleneck layer includes a series of Conv layers, a mixed attention layer and a Conv layer, and the mixed attention layer includes a parallel SE attention layer and an MPCA attention layer; the neck network Neck includes a fifth module and a sixth module; the fifth module includes an Upsample layer, a C2f layer and an Upsample layer connected in series in sequence; the sixth module includes a seventh module, an eighth module and a C2f layer connected in series; the seventh module and the eighth module both include a C2f layer and a CBS layer connected in series; the head network Head includes three parallel Detect layers; wherein the SE attention layer adopts a Squeeze-and-Excitation channel attention mechanism, and the MPCA attention layer adopts a multi-scale position attention mechanism.

7. A multi-parameter based hydropower station maintenance method according to claim 6, characterized in that: The rockfall disaster prediction model is trained, including: The preprocessed historical geological data and historical meteorological data are input into the long short-term memory network, and the initial model parameters of the long short-term memory network are determined by a cross-validation method; the initial model parameters are applied to the long short-term memory network, and the parameters of the long short-term memory network are adjusted using a grid search method to obtain a trained rockfall disaster prediction model.

8. A multi-parameter based hydropower station maintenance method according to claim 6, characterized in that: The method for obtaining the judgment result includes: Input the monitoring data into the image extraction module for processing; divide the monitoring video data into video data with a time length of 1 hour; set the frame rate, and obtain the number of monitoring image data corresponding to the monitoring video data according to the frame rate and the time length; perform image division on the monitoring video data according to the number of monitoring image data to obtain corresponding monitoring image data; Input the monitoring image data into the YOLO model for processing to obtain multiple initial detection frames; calculate the confidence of the multiple initial detection frames, and respectively determine whether the confidence corresponding to the multiple initial detection frames exceeds a preset confidence threshold; if so, retain the corresponding initial detection frame; otherwise, discard the corresponding initial detection frame to obtain a final detection frame; The final detection frame and the monitoring image data are input into the CNN model to determine the corresponding rockfall detection result; according to the rockfall detection result, each final detection frame is traversed and the centroid coordinates of the corresponding internal pixel values ​​are calculated; based on the intersection of each centroid coordinate and the corresponding final detection frame boundary, the spatial position of each rockfall is calculated; according to the spatial position, the number of rocks in each final detection frame is calculated to obtain the total number of rocksfall; according to the total number of rocksfall and the pixel value of the corresponding monitoring image data, the total volume of rocksfall is calculated; Based on the water flow velocity of the hydropower station, the total volume of the falling rocks is processed by the falling rocks judgment module; when the water flow velocity of the hydropower station is greater than a preset water flow threshold, if so, it is determined that there is a risk to the turbine of the hydropower station, and the risk level is determined according to the total volume of the falling rocks; otherwise, it is determined whether the total volume of the falling rocks reaches the preset volume threshold, if so, it is determined that the water inlet of the hydropower station needs to be cleaned, otherwise, it is determined that the water inlet of the hydropower station does not need to be cleaned; wherein, the risk levels include high, medium and low; the judgment result includes the risk level and whether the water inlet needs to be cleaned.

9. A multi-parameter based hydropower station maintenance method according to claim 8, characterized in that: The method of formulating a maintenance plan for the hydropower station according to the judgment result and the prediction result includes: Determine whether any two of the following conditions are met: the prediction result exceeds the first threshold, the risk level is medium or high, and the water inlet of the hydropower station needs to be cleaned; if so, quickly arrange technicians to clean the water inlet, check the status of the turbine of the hydropower station and perform repairs, and check and strengthen the stability of the surrounding protective net; Determine whether the prediction result exceeds the second threshold value and satisfies any one of the following conditions: the risk level is medium or high, and the water inlet of the hydropower station needs to be cleaned; then arrange technicians to check the stability of the surrounding protective nets and check or repair the status of the turbine or clean the water inlet; In other cases, the original maintenance plan is implemented.

10. A multi-parameter based hydropower station maintenance system, used to implement a multi-parameter based hydropower station maintenance method according to any one of claims 1 to 9, characterized in that: It includes a data acquisition module, a data processing module, a state determination module, a rockfall statistics module, a rockfall disaster prediction module and a maintenance plan formulation module; the data processing module includes a data preprocessing submodule and a threshold determination submodule; wherein: A data acquisition module is used to obtain the actual operation data of the hydropower station, historical rockfall event data, monitoring data, geological data and meteorological data; A data preprocessing submodule, for performing data cleaning, feature selection and normalization processing on the historical rockfall event data, the geological data and the meteorological data; A threshold determination submodule, used to calculate the actual power generation efficiency of the hydropower station based on the actual operation data; and determine a preset power generation threshold of the hydropower station; A state determination module, used for determining the current state of the hydropower station based on the preset power generation threshold and the actual power generation efficiency; A rockfall statistics module is used to process the monitoring data through the rockfall judgment model according to the current state of the hydropower station to obtain a corresponding judgment result; A rockfall disaster prediction module is used to process the preprocessed geological data and the preprocessed meteorological data using the trained rockfall disaster prediction model according to the current state of the hydropower station to obtain corresponding prediction results; The maintenance plan formulation module is used to formulate a maintenance plan for the hydropower station based on the judgment result and the prediction result.

Images