Pumped storage unit shaft system attitude prediction method and device

Abstract

The invention relates to the technical field of pumped storage unit health assessment, and particularly provides a pumped storage unit shaft system attitude prediction method and device, and the method comprises the steps: taking the displacement waveform data of a bearing in a pumped storage unit in a current time period as the input of a pre-trained convolutional neural network model based on an attention mechanism and a long-short term memory network; obtaining the curvature of the main shaft in the pumped storage unit in a prediction time period output by a pre-trained convolutional neural network model based on the attention mechanism and the long-short term memory network; and analyzing the shafting attitude of the pumped storage unit in the prediction time period based on the curvature of the main shaft in the pumped storage unit in the prediction time period. According to the technical scheme provided by the invention, the distortion degree of the shafting attitude of the pumping and storage unit is evaluated by the quantitative index, the real-time state of the shafting attitude of the pumping and storage unit can be intuitively displayed, and the shafting attitude of the pumping and storage unit under different working conditions can be accurately predicted.

Description

Technical Field

[0001] The present invention relates to the technical field of health assessment of pumped storage units, and in particular to a method and device for predicting the attitude of an axis system of a pumped storage unit. Background Art

[0002] As the proportion of new energy sources such as wind power and photovoltaic power generation increases, the safety and stability of the power grid faces new challenges. As an important regulator, the functional positioning, operation mode, and operation intensity of pumped storage units have also undergone great changes. The start-stop frequency of pumped storage units has doubled, and the units often need to operate at the boundary of the stable zone.

[0003] The shaft system of the pumped storage unit is subject to hydraulic shock, mechanical failure and electromagnetic imbalance during frequent start-stop and working condition switching, which may induce various faults and accidents such as abnormal vibration, structural fatigue, electrical failure, and damage to the operation mode of the unit, which are extremely harmful. The main shaft of the pumped storage unit will be distorted after long-term operation, which will seriously affect the operating safety and reliability of the pumped storage unit. The axis of the pumped storage unit is the static performance of the main shaft. The axis adjustment is an indispensable process in the installation and maintenance management of the unit. The axis swing is a key indicator for identifying the quality of the unit axis. The axis trajectory of the pumped storage unit is the vibration trajectory of the main shaft center and the dynamic performance of the main shaft when rotating. By observing and monitoring the axis trajectory, the operating status of the equipment can be obtained vividly and intuitively. At present, experts and scholars have conducted in-depth research on the axis trajectory, including axis trajectory shape recognition, axis trajectory purification, axis trajectory feature extraction and fault diagnosis based on axis trajectory, but there is little research on the main shaft posture under the rotating state. Summary of the invention

[0004] In order to overcome the above-mentioned defects, the present invention proposes a method and device for predicting the shaft system posture of a pumped storage unit.

[0005] In a first aspect, a method for predicting the shaft system attitude of a pumped storage unit is provided, the method comprising:

[0006] The displacement waveform data of the bearing in the pumped storage unit in the current period is used as the input of the pre-trained convolutional neural network model based on the attention mechanism and the long short-term memory network, and the curvature of the main shaft in the pumped storage unit in the predicted period is obtained as the output of the pre-trained convolutional neural network model based on the attention mechanism and the long short-term memory network;

[0007] The shaft system posture of the pumped-storage unit within the prediction period is analyzed based on the curvature of the main shaft in the pumped-storage unit within the prediction period.

[0008] Preferably, the displacement waveform data of the bearing includes: lateral displacement waveform data of the upper guide bearing, longitudinal displacement waveform data of the upper guide bearing, lateral displacement waveform data of the lower guide bearing, longitudinal displacement waveform data of the lower guide bearing, lateral displacement waveform data of the water guide bearing, and longitudinal displacement waveform data of the water guide bearing.

[0009] Preferably, the displacement waveform data of the bearing is acquired by an eddy current displacement sensor.

[0010] Furthermore, the training process of the pre-trained convolutional neural network model based on the attention mechanism and the long short-term memory network includes:

[0011] Obtaining displacement waveform data of bearings in a pumped-storage unit in a historical period, and calculating standardized data corresponding to the displacement waveform data of the bearings in the historical period;

[0012] determining the curvature of the main shaft in the hydroelectric storage unit during the historical period based on the standardized data;

[0013] Constructing training data using displacement waveform data of bearings in the pumped storage unit in the historical period and curvature of the main shaft in the water storage unit in the historical period;

[0014] The training data is used to train the initial convolutional neural network model based on the attention mechanism and the long short-term memory network to obtain the pre-trained convolutional neural network model based on the attention mechanism and the long short-term memory network.

[0015] Furthermore, the standardized data corresponding to the displacement waveform data of the bearing in the historical period is as follows:

[0016] x s =x-μ

[0017] In the above formula, x s is the standardized data corresponding to the displacement waveform data of the bearing in the historical period, x is the displacement waveform data of the bearing in the historical period, and μ is the mean value of the displacement waveform data of the bearing in the historical period.

[0018] Further, the determining the curvature of the main shaft of the water storage unit in the historical period based on the standardized data includes:

[0019] A three-dimensional coordinate system is constructed with the water guide bearing as the origin, the main axis direction as the z-axis, the upper guide bearing transverse direction, the lower guide bearing transverse direction or the water guide bearing transverse direction as the x-axis, and the upper guide bearing longitudinal direction, the lower guide bearing longitudinal direction or the water guide bearing longitudinal direction as the y-axis;

[0020] The lateral displacement waveform data of the upper guide bearing in the pumped storage unit is used as the x-axis coordinate value of the upper guide bearing in the three-dimensional coordinate system, the longitudinal displacement waveform data of the upper guide bearing is used as the y-axis coordinate value of the upper guide bearing in the three-dimensional coordinate system, and the relative distance between the upper guide bearing and the water guide bearing is used as the z-axis coordinate value of the upper guide bearing in the three-dimensional coordinate system, to obtain the three-dimensional coordinates of the upper guide bearing in the pumped storage unit at the historical moment;

[0021] The lateral displacement waveform data of the lower guide bearing in the pumped storage unit is used as the x-axis coordinate value of the lower guide bearing in the three-dimensional coordinate system, the longitudinal displacement waveform data of the lower guide bearing is used as the y-axis coordinate value of the lower guide bearing in the three-dimensional coordinate system, and the relative distance between the lower guide bearing and the water guide bearing is used as the z-axis coordinate value of the lower guide bearing in the three-dimensional coordinate system, to obtain the three-dimensional coordinates of the lower guide bearing in the pumped storage unit at the historical moment;

[0022] The lateral displacement waveform data of the water-guide bearing in the pumped-storage unit is used as the x-axis coordinate value of the water-guide bearing in the three-dimensional coordinate system, the longitudinal displacement waveform data of the water-guide bearing is used as the y-axis coordinate value of the water-guide bearing in the three-dimensional coordinate system, and 0 is used as the z-axis coordinate value of the water-guide bearing in the three-dimensional coordinate system, to obtain the three-dimensional coordinates of the water-guide bearing in the pumped-storage unit at the historical moment;

[0023] The curvature of the main shaft in the pumped storage unit at the historical moment is determined based on the three-dimensional coordinates of the upper, lower and water-guided bearings in the pumped storage unit at the historical moment.

[0024] Furthermore, the curvature of the main shaft in the water storage unit at the historical moment is as follows:

[0025]

[0026] In the above formula, S is the curvature of the main shaft in the water storage unit at the historical moment, s is the relative distance between the upper guide bearing and the water guide bearing, s′ is the vertical distance between the lower guide bearing and the straight line a, and the endpoints of the straight line a are the three-dimensional coordinates of the upper guide bearing and the three-dimensional coordinates of the water guide bearing, respectively.

[0027] Furthermore, the relative distance between the upper guide bearing and the water guide bearing is as follows:

[0028]

[0029] In the above formula, A 上 , B 上 and C 上 A is the three-dimensional coordinate x, y and z axis coordinate value of the upper guide bearing, 水 , B 水 and C 水 are the three-dimensional x, y and z axis coordinate values ​​of the water guide bearing.

[0030] Furthermore, the vertical distance between the lower guide bearing and the straight line a is as follows:

[0031]

[0032] In the above formula, P 上 is the three-dimensional coordinate of the upper guide bearing, P 下 is the three-dimensional coordinate of the lower guide bearing, P 水 is the three-dimensional coordinate of the water-guided bearing.

[0033] Preferably, the analysis of the shaft system posture of the pumped-storage unit within the prediction period based on the curvature of the main shaft in the pumped-storage unit within the prediction period includes:

[0034] When the curvature of the main shaft in the pumped-storage unit in the prediction period is a positive value, the main shaft in the pumped-storage unit is in a convex posture. When the curvature of the main shaft in the pumped-storage unit in the prediction period is a negative value, the main shaft in the pumped-storage unit is in a concave posture.

[0035] Furthermore, the initial convolutional neural network of the convolutional neural network model based on the attention mechanism and the long short-term memory network includes:

[0036] Layer 1: Fully connected layer, with 6 input features and 30 output features;

[0037] Layer 2: 2D convolutional layer, with 1 input channel, 64 output channels, stride 1, kernel size 3, and same padding;

[0038] Layer 3: ReLU activation function;

[0039] Layer 4: 2D max pooling layer with pooling size of 2, stride of 1, and same padding;

[0040] Layer 5: 2D dropout layer with dropout probability set to 0.3;

[0041] Layer 6: Long short-term memory network, input size is 64*30, hidden state size is 256, number of recurrent network layers is 1, time_major is set to False;

[0042] Layer 7: LSTM network, input size is 256, hidden state size is 128, number of recurrent layers is 1, time_major is set to False;

[0043] Layer 8: Long short-term memory network, input size is 128, hidden state size is 64, number of recurrent network layers is 1, time_major is set to False;

[0044] Layer 9: Fully connected layer, with 64 input features and 32 output features;

[0045] Layer 10: ReLU activation function;

[0046] Layer 11: Fully connected layer, with an input feature of 32 and an output feature of 1.

[0047] Furthermore, the attention mechanism of the initial convolutional neural network model based on the attention mechanism and the long short-term memory network includes: a channel attention model and a spatial attention model;

[0048] The channel attention model includes:

[0049] Layer 1: 2D adaptive average pooling layer, the output feature map size is 1;

[0050] Layer 2: 2D adaptive max pooling layer with an output feature map size of 1;

[0051] Layer 3: 2D convolutional layer;

[0052] Layer 4: activation function ReLU;

[0053] Layer 5: 2D convolutional layer;

[0054] Layer 6: activation function Sigmoid;

[0055] The spatial attention model includes:

[0056] Layer 1: 2D convolutional layer;

[0057] Layer 2: Activation function Sigmoid.

[0058] Furthermore, the channel attention model is placed in the third layer of the convolutional neural network: after the ReLU activation function, and the spatial attention model is placed after the channel attention model.

[0059] In a second aspect, a device for predicting the shaft system attitude of a pumped storage unit is provided, wherein the device for predicting the shaft system attitude of a pumped storage unit comprises:

[0060] The first analysis module is used to use the displacement waveform data of the bearing in the pumped-storage unit in the current period as the input of the pre-trained convolutional neural network model based on the attention mechanism and the long short-term memory network, and obtain the curvature of the main shaft in the pumped-storage unit in the predicted period output by the pre-trained convolutional neural network model based on the attention mechanism and the long short-term memory network;

[0061] The second analysis module is used to analyze the shaft system posture of the pumped-storage unit within the prediction period based on the curvature of the main shaft in the pumped-storage unit within the prediction period.

[0062] Preferably, the displacement waveform data of the bearing includes: lateral displacement waveform data of the upper guide bearing, longitudinal displacement waveform data of the upper guide bearing, lateral displacement waveform data of the lower guide bearing, longitudinal displacement waveform data of the lower guide bearing, lateral displacement waveform data of the water guide bearing, and longitudinal displacement waveform data of the water guide bearing.

[0063] Preferably, the displacement waveform data of the bearing is acquired by an eddy current displacement sensor.

[0064] Furthermore, the training process of the pre-trained convolutional neural network model based on the attention mechanism and the long short-term memory network includes:

[0065] Obtaining displacement waveform data of bearings in a pumped-storage unit in a historical period, and calculating standardized data corresponding to the displacement waveform data of the bearings in the historical period;

[0066] determining the curvature of the main shaft in the hydroelectric storage unit during the historical period based on the standardized data;

[0067] Constructing training data using displacement waveform data of bearings in the pumped storage unit in the historical period and curvature of the main shaft in the water storage unit in the historical period;

[0068] The training data is used to train the initial convolutional neural network model based on the attention mechanism and the long short-term memory network to obtain the pre-trained convolutional neural network model based on the attention mechanism and the long short-term memory network.

[0069] Furthermore, the standardized data corresponding to the displacement waveform data of the bearing in the historical period is as follows:

[0070] x s =x-μ

[0071] In the above formula, x s is the standardized data corresponding to the displacement waveform data of the bearing in the historical period, x is the displacement waveform data of the bearing in the historical period, and μ is the mean value of the displacement waveform data of the bearing in the historical period.

[0072] Further, the determining the curvature of the main shaft of the water storage unit in the historical period based on the standardized data includes:

[0073] A three-dimensional coordinate system is constructed with the water guide bearing as the origin, the main axis direction as the z-axis, the upper guide bearing transverse direction, the lower guide bearing transverse direction or the water guide bearing transverse direction as the x-axis, and the upper guide bearing longitudinal direction, the lower guide bearing longitudinal direction or the water guide bearing longitudinal direction as the y-axis;

[0074] The lateral displacement waveform data of the upper guide bearing in the pumped storage unit is used as the x-axis coordinate value of the upper guide bearing in the three-dimensional coordinate system, the longitudinal displacement waveform data of the upper guide bearing is used as the y-axis coordinate value of the upper guide bearing in the three-dimensional coordinate system, and the relative distance between the upper guide bearing and the water guide bearing is used as the z-axis coordinate value of the upper guide bearing in the three-dimensional coordinate system, to obtain the three-dimensional coordinates of the upper guide bearing in the pumped storage unit at the historical moment;

[0075] The lateral displacement waveform data of the lower guide bearing in the pumped storage unit is used as the x-axis coordinate value of the lower guide bearing in the three-dimensional coordinate system, the longitudinal displacement waveform data of the lower guide bearing is used as the y-axis coordinate value of the lower guide bearing in the three-dimensional coordinate system, and the relative distance between the lower guide bearing and the water guide bearing is used as the z-axis coordinate value of the lower guide bearing in the three-dimensional coordinate system, to obtain the three-dimensional coordinates of the lower guide bearing in the pumped storage unit at the historical moment;

[0076] The lateral displacement waveform data of the water-guide bearing in the pumped-storage unit is used as the x-axis coordinate value of the water-guide bearing in the three-dimensional coordinate system, the longitudinal displacement waveform data of the water-guide bearing is used as the y-axis coordinate value of the water-guide bearing in the three-dimensional coordinate system, and 0 is used as the z-axis coordinate value of the water-guide bearing in the three-dimensional coordinate system, to obtain the three-dimensional coordinates of the water-guide bearing in the pumped-storage unit at the historical moment;

[0077] The curvature of the main shaft in the pumped storage unit at the historical moment is determined based on the three-dimensional coordinates of the upper, lower and water-guided bearings in the pumped storage unit at the historical moment.

[0078] Furthermore, the curvature of the main shaft in the water storage unit at the historical moment is as follows:

[0079]

[0080] In the above formula, S is the curvature of the main shaft in the water storage unit at the historical moment, s is the relative distance between the upper guide bearing and the water guide bearing, s′ is the vertical distance between the lower guide bearing and the straight line a, and the endpoints of the straight line a are the three-dimensional coordinates of the upper guide bearing and the three-dimensional coordinates of the water guide bearing, respectively.

[0081] Furthermore, the relative distance between the upper guide bearing and the water guide bearing is as follows:

[0082]

[0083] In the above formula, A 上 , B 上 and C 上 A is the three-dimensional coordinate x, y and z axis coordinate value of the upper guide bearing, 水 , B 水 and C 水 are the three-dimensional x, y and z axis coordinate values ​​of the water guide bearing.

[0084] Furthermore, the vertical distance between the lower guide bearing and the straight line a is as follows:

[0085]

[0086] In the above formula, P 上 is the three-dimensional coordinate of the upper guide bearing, P 下 is the three-dimensional coordinate of the lower guide bearing, P 水 is the three-dimensional coordinate of the water-guided bearing.

[0087] Preferably, the second analysis module is specifically used for:

[0088] When the curvature of the main shaft in the pumped-storage unit in the prediction period is a positive value, the main shaft in the pumped-storage unit is in a convex posture. When the curvature of the main shaft in the pumped-storage unit in the prediction period is a negative value, the main shaft in the pumped-storage unit is in a concave posture.

[0089] Furthermore, the initial convolutional neural network of the convolutional neural network model based on the attention mechanism and the long short-term memory network includes:

[0090] Layer 1: Fully connected layer, with 6 input features and 30 output features;

[0091] Layer 2: 2D convolutional layer, with 1 input channel, 64 output channels, stride 1, kernel size 3, and same padding;

[0092] Layer 3: ReLU activation function;

[0093] Layer 4: 2D max pooling layer with pooling size of 2, stride of 1, and same padding;

[0094] Layer 5: 2D dropout layer with dropout probability set to 0.3;

[0095] Layer 6: Long short-term memory network, input size is 64*30, hidden state size is 256, number of recurrent network layers is 1, time_major is set to False;

[0096] Layer 7: LSTM network, input size is 256, hidden state size is 128, number of recurrent layers is 1, time_major is set to False;

[0097] Layer 8: Long short-term memory network, input size is 128, hidden state size is 64, number of recurrent network layers is 1, time_major is set to False;

[0098] Layer 9: Fully connected layer, with 64 input features and 32 output features;

[0099] Layer 10: ReLU activation function;

[0100] Layer 11: Fully connected layer, with an input feature of 32 and an output feature of 1.

[0101] Furthermore, the attention mechanism of the initial convolutional neural network model based on the attention mechanism and the long short-term memory network includes: a channel attention model and a spatial attention model;

[0102] The channel attention model includes:

[0103] Layer 1: 2D adaptive average pooling layer, the output feature map size is 1;

[0104] Layer 2: 2D adaptive max pooling layer with an output feature map size of 1;

[0105] Layer 3: 2D convolutional layer;

[0106] Layer 4: activation function ReLU;

[0107] Layer 5: 2D convolutional layer;

[0108] Layer 6: activation function Sigmoid;

[0109] The spatial attention model includes:

[0110] Layer 1: 2D convolutional layer;

[0111] Layer 2: Activation function Sigmoid.

[0112] Furthermore, the channel attention model is placed in the third layer of the convolutional neural network: after the ReLU activation function, and the spatial attention model is placed after the channel attention model.

[0113] In a third aspect, a computer device is provided, comprising: one or more processors;

[0114] The processor is used to store one or more programs;

[0115] When the one or more programs are executed by the one or more processors, the method for predicting the shaft system attitude of the pumped-storage unit is implemented.

[0116] In a fourth aspect, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed, the method for predicting the shaft system attitude of the pumped-storage unit is implemented.

[0117] The above one or more technical solutions of the present invention have at least one or more of the following beneficial effects:

[0118] The present invention provides a method and device for predicting the shaft attitude of a pumped storage unit, comprising: using the displacement waveform data of the bearing in the pumped storage unit in the current period as the input of a pre-trained convolutional neural network model based on an attention mechanism and a long short-term memory network, obtaining the curvature of the main shaft in the pumped storage unit in the prediction period output by the pre-trained convolutional neural network model based on an attention mechanism and a long short-term memory network; and analyzing the shaft attitude of the pumped storage unit in the prediction period based on the curvature of the main shaft in the pumped storage unit in the prediction period. The technical solution provided by the present invention evaluates the distortion degree of the shaft attitude of the pumped storage unit with quantitative indicators, can intuitively display the real-time state of the shaft attitude of the pumped storage unit, and accurately predict the shaft attitude of the pumped storage unit under different working conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0119] Figure 1 It is a schematic flow chart of the main steps of the method for predicting the shaft system posture of a pumped storage unit according to an embodiment of the present invention;

[0120] Figure 2 is a comparison diagram of the axis trajectory of the sensor before and after adjustment of the embodiment of the present invention;

[0121] Figure 3 Schematic diagram of the shaft system posture of the pumped storage unit according to an embodiment of the present invention;

[0122] Figure 4 : is a comparison diagram between the actual lateral value and the predicted value of the lower guide swing according to an embodiment of the present invention;

[0123] Figure 5 Schematic diagram of predicted axis system attitude according to an embodiment of the present invention. DETAILED DESCRIPTION

[0124] The specific implementation modes of the present invention will be further described in detail below in conjunction with the accompanying drawings.

[0125] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution 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 part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0126] As disclosed in the background technology, as the proportion of wind power, photovoltaic and other new energy power generation increases, the safety and stability of the power grid faces new challenges. As an important regulator, the functional positioning, operation mode and operation intensity of the pumped storage unit have also undergone great changes. The start and stop frequency of the pumped storage unit has increased exponentially, and the unit often needs to operate at the boundary of the stable zone.

[0127] The shaft system of the pumped storage unit is subject to hydraulic shock, mechanical failure and electromagnetic imbalance during frequent start-stop and working condition switching, which may induce various faults and accidents such as abnormal vibration, structural fatigue, electrical failure, and damage to the operation mode of the unit, which are extremely harmful. The main shaft of the pumped storage unit will be distorted after long-term operation, which will seriously affect the operating safety and reliability of the pumped storage unit. The axis of the pumped storage unit is the static performance of the main shaft. The axis adjustment is an indispensable process in the installation and maintenance management of the unit. The axis swing is a key indicator for identifying the quality of the unit axis. The axis trajectory of the pumped storage unit is the vibration trajectory of the main shaft center and the dynamic performance of the main shaft when rotating. By observing and monitoring the axis trajectory, the operating status of the equipment can be obtained vividly and intuitively. At present, experts and scholars have conducted in-depth research on the axis trajectory, including axis trajectory shape recognition, axis trajectory purification, axis trajectory feature extraction and fault diagnosis based on axis trajectory, but there is little research on the main shaft posture under the rotating state.

[0128] In order to improve the above-mentioned problems, the present invention relates to the technical field of health assessment of pumped storage units, and specifically provides a method and device for predicting the shaft system attitude of a pumped storage unit, including: using the displacement waveform data of the bearings in the pumped storage unit in the current period as the input of a pre-trained convolutional neural network model based on an attention mechanism and a long short-term memory network, and obtaining the curvature of the main shaft in the pumped storage unit in the prediction period output by the pre-trained convolutional neural network model based on an attention mechanism and a long short-term memory network; analyzing the shaft system attitude of the pumped storage unit in the prediction period based on the curvature of the main shaft in the pumped storage unit in the prediction period. The technical solution provided by the present invention evaluates the degree of distortion of the shaft system attitude of the pumped storage unit with quantitative indicators, can intuitively display the real-time state of the shaft system attitude of the pumped storage unit, and accurately predict the shaft system attitude of the pumped storage unit under different working conditions. The above-mentioned scheme is elaborated in detail below.

[0129] Example 1

[0130] See attached Figure 1 , Figure 1 FIG. 1 is a schematic flow chart of the main steps of a method for predicting the shaft system posture of a pumped storage unit according to an embodiment of the present invention. Figure 1 As shown, the method for predicting the shaft system attitude of a pumped storage unit in the embodiment of the present invention mainly includes the following steps:

[0131] Step S101: using the displacement waveform data of the bearing in the pumped-storage unit in the current period as the input of the pre-trained convolutional neural network model based on the attention mechanism and the long short-term memory network, and obtaining the curvature of the main shaft in the pumped-storage unit in the predicted period output by the pre-trained convolutional neural network model based on the attention mechanism and the long short-term memory network;

[0132] Step S102: analyzing the shaft system posture of the pumped-storage unit within the prediction period based on the curvature of the main shaft in the pumped-storage unit within the prediction period.

[0133] In this embodiment, the displacement waveform data of the bearing includes: the lateral displacement waveform data of the upper guide bearing, the longitudinal displacement waveform data of the upper guide bearing, the lateral displacement waveform data of the lower guide bearing, the longitudinal displacement waveform data of the lower guide bearing, the lateral displacement waveform data of the water guide bearing, and the longitudinal displacement waveform data of the water guide bearing.

[0134] Wherein, the displacement waveform data of the bearing is obtained by an eddy current displacement sensor.

[0135] In one embodiment, the training process of the pre-trained convolutional neural network model based on the attention mechanism and the long short-term memory network includes:

[0136] Obtaining displacement waveform data of bearings in a pumped-storage unit in a historical period, and calculating standardized data corresponding to the displacement waveform data of the bearings in the historical period;

[0137] determining the curvature of the main shaft in the hydroelectric storage unit during the historical period based on the standardized data;

[0138] Constructing training data using displacement waveform data of bearings in the pumped storage unit in the historical period and curvature of the main shaft in the water storage unit in the historical period;

[0139] The training data is used to train the initial convolutional neural network model based on the attention mechanism and the long short-term memory network to obtain the pre-trained convolutional neural network model based on the attention mechanism and the long short-term memory network.

[0140] In one embodiment, since the distance between the eddy current displacement sensor and the main shaft will affect the relative positions of the three plane axis tracks, the axis track of the guide bearing is taken as an example. Figure 2 . Adjust the eddy current displacement sensors in the X and Y directions of the upper guide bearing close to the main shaft. It can be seen that the shape of the axis trajectory has not changed significantly, but the position of the axis trajectory has shifted significantly. The results of judging the axis posture by these two axis trajectories will be completely different. In order to remove this influencing factor, the waveform data of the six measuring points are standardized using the centralization method. The standardized data corresponding to the displacement waveform data of the bearing during the historical period are as follows:

[0141] x s =x-μ

[0142] In the above formula, x s is the standardized data corresponding to the displacement waveform data of the bearing in the historical period, x is the displacement waveform data of the bearing in the historical period, and μ is the mean value of the displacement waveform data of the bearing in the historical period.

[0143] In one embodiment, the determining the curvature of the main shaft in the water storage unit in the historical period based on the standardized data includes:

[0144] A three-dimensional coordinate system is constructed with the water guide bearing as the origin, the main axis direction as the z-axis, the upper guide bearing transverse direction, the lower guide bearing transverse direction or the water guide bearing transverse direction as the x-axis, and the upper guide bearing longitudinal direction, the lower guide bearing longitudinal direction or the water guide bearing longitudinal direction as the y-axis;

[0145] The lateral displacement waveform data of the upper guide bearing in the pumped storage unit is used as the x-axis coordinate value of the upper guide bearing in the three-dimensional coordinate system, the longitudinal displacement waveform data of the upper guide bearing is used as the y-axis coordinate value of the upper guide bearing in the three-dimensional coordinate system, and the relative distance between the upper guide bearing and the water guide bearing is used as the z-axis coordinate value of the upper guide bearing in the three-dimensional coordinate system, to obtain the three-dimensional coordinates of the upper guide bearing in the pumped storage unit at the historical moment;

[0146] The lateral displacement waveform data of the lower guide bearing in the pumped storage unit is used as the x-axis coordinate value of the lower guide bearing in the three-dimensional coordinate system, the longitudinal displacement waveform data of the lower guide bearing is used as the y-axis coordinate value of the lower guide bearing in the three-dimensional coordinate system, and the relative distance between the lower guide bearing and the water guide bearing is used as the z-axis coordinate value of the lower guide bearing in the three-dimensional coordinate system, to obtain the three-dimensional coordinates of the lower guide bearing in the pumped storage unit at the historical moment;

[0147] The lateral displacement waveform data of the water-guide bearing in the pumped-storage unit is used as the x-axis coordinate value of the water-guide bearing in the three-dimensional coordinate system, the longitudinal displacement waveform data of the water-guide bearing is used as the y-axis coordinate value of the water-guide bearing in the three-dimensional coordinate system, and 0 is used as the z-axis coordinate value of the water-guide bearing in the three-dimensional coordinate system, to obtain the three-dimensional coordinates of the water-guide bearing in the pumped-storage unit at the historical moment;

[0148] The curvature of the main shaft in the pumped storage unit at the historical moment is determined based on the three-dimensional coordinates of the upper, lower and water-guided bearings in the pumped storage unit at the historical moment.

[0149] In a specific implementation, in the three-dimensional graph, the first data point of the upper guide axis trajectory, the lower guide axis trajectory and the water guide axis trajectory is obtained and connected to form an axis attitude line. Then the data points are obtained in sequence and connected, and finally all the axis attitude lines form an axis attitude surface. Since the main shaft will be slightly twisted and deformed when the pumped storage unit rotates, two basic axis attitudes will be generated. See the schematic diagram Figure 3 .

[0150] In one embodiment, the curvature of the main shaft in the water storage unit at the historical moment is as follows:

[0151]

[0152] In the above formula, S is the curvature of the main shaft in the water storage unit at the historical moment, s is the relative distance between the upper guide bearing and the water guide bearing, s′ is the vertical distance between the lower guide bearing and the straight line a, and the endpoints of the straight line a are the three-dimensional coordinates of the upper guide bearing and the three-dimensional coordinates of the water guide bearing, respectively.

[0153] In one embodiment, the relative distance between the upper guide bearing and the water guide bearing is as follows:

[0154]

[0155] In the above formula, A 上 , B 上 and C 上 A is the three-dimensional coordinate x, y and z axis coordinate value of the upper guide bearing, 水 , B 水 and C 水 are the three-dimensional x, y and z axis coordinate values ​​of the water guide bearing.

[0156] In one embodiment, the vertical distance between the lower guide bearing and the straight line a is as follows:

[0157]

[0158] In the above formula, P 上 is the three-dimensional coordinate of the upper guide bearing, P 下 is the three-dimensional coordinate of the lower guide bearing, P 水 is the three-dimensional coordinate of the water-guided bearing.

[0159] In this embodiment, the shaft system posture of the pumped-storage unit within the prediction period is analyzed based on the curvature of the main shaft in the pumped-storage unit within the prediction period, including:

[0160] When the curvature of the main shaft in the pumped-storage unit during the prediction period is a positive value, the main shaft in the pumped-storage unit is in a convex posture, and the larger the value, the more serious the convexity of the shaft. When the curvature of the main shaft in the pumped-storage unit during the prediction period is a negative value, the main shaft in the pumped-storage unit is in a concave posture, and the smaller the value, the more serious the concaveness of the shaft.

[0161] In one embodiment, the initial convolutional neural network based on the convolutional neural network model of the attention mechanism and the long short-term memory network includes:

[0162] Layer 1: Fully connected layer, with 6 input features and 30 output features;

[0163] Layer 2: 2D convolutional layer, with 1 input channel, 64 output channels, stride 1, kernel size 3, and same padding;

[0164] Layer 3: ReLU activation function;

[0165] Layer 4: 2D max pooling layer with pooling size of 2, stride of 1, and same padding;

[0166] Layer 5: 2D dropout layer with dropout probability set to 0.3;

[0167] Layer 6: Long short-term memory network, input size is 64*30, hidden state size is 256, number of recurrent network layers is 1, time_major is set to False;

[0168] Layer 7: LSTM network, input size is 256, hidden state size is 128, number of recurrent layers is 1, time_major is set to False;

[0169] Layer 8: Long short-term memory network, input size is 128, hidden state size is 64, number of recurrent network layers is 1, time_major is set to False;

[0170] Layer 9: Fully connected layer, with 64 input features and 32 output features;

[0171] Layer 10: ReLU activation function;

[0172] Layer 11: Fully connected layer, with an input feature of 32 and an output feature of 1.

[0173] In one embodiment, the attention mechanism of the initial convolutional neural network model based on the attention mechanism and the long short-term memory network includes: a channel attention model and a spatial attention model;

[0174] The channel attention model includes:

[0175] Layer 1: 2D adaptive average pooling layer, the output feature map size is 1;

[0176] Layer 2: 2D adaptive max pooling layer with an output feature map size of 1;

[0177] Layer 3: 2D convolutional layer;

[0178] Layer 4: activation function ReLU;

[0179] Layer 5: 2D convolutional layer;

[0180] Layer 6: activation function Sigmoid;

[0181] The spatial attention model includes:

[0182] Layer 1: 2D convolutional layer;

[0183] Layer 2: Activation function Sigmoid.

[0184] In one embodiment, the channel attention model is placed in the third layer of the convolutional neural network: after the ReLU activation function, and the spatial attention model is placed after the channel attention model.

[0185] In a specific implementation, model training is performed based on PaddlePaddle2.3.2, the initial learning rate is set to 0.001, the number of training rounds is 200, the data size of each round is 32, and the optimizer first uses CosineAnnealingDecay to dynamically adjust the learning rate, and then outputs the learning rate to the Adam optimizer. The exponential decay rate of the first-order moment estimate of the Adam optimizer is 0.9, and the exponential decay rate of the second-order moment estimate is 0.999.

[0186] In a specific implementation, taking a pumped storage power station as an example, the model input data is the waveform data and curvature S of 6 measuring points, including upper guide swing X, upper guide swing Y, lower guide swing X, lower guide swing Y, water guide swing X, and water guide swing Y. Each measuring point has 2176 values, and 70% is selected as the training set and 30% is selected as the test set. Taking each swing data as the feature data x and the curvature as the label y, the size of train_x in the training set is (1483, 6), and the size of train_y is (1483, 1). The size of test_x in the test set is (693, 6), and the size of test_y is (693, 1). The model parameters are shown in Table 1.

[0187] Table 1

[0188] Layer(type) Input Shape Output Shape Param Linear [[1483,1,40,6]] [1483,1,40,30] 210 Conv2D [[1483,1,40,30]] [1483,64,40,30] 640 ReLU [[1483,64,40,30]] [1483,64,40,30] 0 AdaptiveAvgPool2D [[1483,64,40,30]] [1483,64,1,1] 0 Conv2D [[1483,64,1,1]] [1483,8,1,1] 520 ReLU [[1483,8,1,1]] [1483,8,1,1] 0 Conv2D [[1483,8,1,1]] [1483,64,1,1] 576 AdaptiveMaxPool2D [[1483,64,40,30]] [1483,64,1,1] 0 Sigmoid [[1483,64,1,1]] [1483,64,1,1] 0 ChannelAttention [[1483,64,40,30]] [1483,64,1,1] 0 Conv2D [[1483,2,40,30]] [1483,1,40,30] 99 Sigmoid [[1483,1,40,30]] [1483,1,40,30] 0 SpatialAttention [[1483,64,40,30]] [1483,1,40,30] 0 CBAM [[1483,64,40,30]] [1483,64,40,30] 0 MaxPool2D [[1483,64,40,30]] [1483,64,40,30] 0 Dropout2D [[1483,64,40,30]] [1483,64,40,30] 0 LSTM [[1483,40,1920]] [[1483,40,256],[[1,1483,256],[1,1483,256]]] 2230272 LSTM [[1483,40,256]] [[1483,40,128],[[1,1483,128],[1,1483,128]]] 197632 LSTM [[1483,40,128]] [[1483,40,64],[[1,1483,64],[1,1483,64]]] 49664 Linear [[1483,64]] [1483,32] 2080 ReLU [[1483,32]] [1483,32] 0 Linear [[1483,32]] [1483,1] 33

[0189] Among them, Layer(type) is the layer (form), Input Shape is the input size, Output Shape is the output size, Param is the parameter, Linear is the fully connected layer, Conv2D is the two-dimensional convolutional layer, ReLU is the activation function, AdaptiveAvgPool2D is the two-dimensional adaptive average pooling layer, AdaptiveMaxPool2D is the two-dimensional adaptive maximum pooling layer, Sigmoid is the activation function, ChannelAttention is the channel attention, SpatialAttention is the spatial attention, CBAM is the attention mechanism, MaxPool2D is the two-dimensional maximum pooling layer, Dropout2D is the two-dimensional Dropout layer, and LSTM is the long short-term memory network.

[0190] Take the swing X as an example, the comparison between the actual value and the predicted value of the measuring point is shown in the figure below. Figure 4The same method can be used to obtain the predicted values ​​of upper guide swing X, upper guide swing Y, lower guide swing Y, water guide swing X, and water guide swing Y. All the predicted values ​​are calculated using the present invention to obtain the predicted axis system attitude. Figure 5 , the figure shows the axis system posture when the main axis rotates one circle.

[0191] Example 2

[0192] Based on the same inventive concept, the present invention also provides a pumped storage unit shaft system attitude prediction device, the pumped storage unit shaft system attitude prediction device comprising:

[0193] The first analysis module is used to use the displacement waveform data of the bearing in the pumped-storage unit in the current period as the input of the pre-trained convolutional neural network model based on the attention mechanism and the long short-term memory network, and obtain the curvature of the main shaft in the pumped-storage unit in the predicted period output by the pre-trained convolutional neural network model based on the attention mechanism and the long short-term memory network;

[0194] The second analysis module is used to analyze the shaft system posture of the pumped-storage unit within the prediction period based on the curvature of the main shaft in the pumped-storage unit within the prediction period.

[0195] Preferably, the displacement waveform data of the bearing includes: lateral displacement waveform data of the upper guide bearing, longitudinal displacement waveform data of the upper guide bearing, lateral displacement waveform data of the lower guide bearing, longitudinal displacement waveform data of the lower guide bearing, lateral displacement waveform data of the water guide bearing, and longitudinal displacement waveform data of the water guide bearing.

[0196] Preferably, the displacement waveform data of the bearing is acquired by an eddy current displacement sensor.

[0197] Furthermore, the training process of the pre-trained convolutional neural network model based on the attention mechanism and the long short-term memory network includes:

[0198] Obtaining displacement waveform data of bearings in a pumped-storage unit in a historical period, and calculating standardized data corresponding to the displacement waveform data of the bearings in the historical period;

[0199] determining the curvature of the main shaft in the hydroelectric storage unit during the historical period based on the standardized data;

[0200] Constructing training data using displacement waveform data of bearings in the pumped storage unit in the historical period and curvature of the main shaft in the water storage unit in the historical period;

[0201] The training data is used to train the initial convolutional neural network model based on the attention mechanism and the long short-term memory network to obtain the pre-trained convolutional neural network model based on the attention mechanism and the long short-term memory network.

[0202] Furthermore, the standardized data corresponding to the displacement waveform data of the bearing in the historical period is as follows:

[0203] x s =x-μ

[0204] In the above formula, x s is the standardized data corresponding to the displacement waveform data of the bearing in the historical period, x is the displacement waveform data of the bearing in the historical period, and μ is the mean value of the displacement waveform data of the bearing in the historical period.

[0205] Further, the determining the curvature of the main shaft of the water storage unit in the historical period based on the standardized data includes:

[0206] A three-dimensional coordinate system is constructed with the water guide bearing as the origin, the main axis direction as the z-axis, the upper guide bearing transverse direction, the lower guide bearing transverse direction or the water guide bearing transverse direction as the x-axis, and the upper guide bearing longitudinal direction, the lower guide bearing longitudinal direction or the water guide bearing longitudinal direction as the y-axis;

[0207] The lateral displacement waveform data of the upper guide bearing in the pumped storage unit is used as the x-axis coordinate value of the upper guide bearing in the three-dimensional coordinate system, the longitudinal displacement waveform data of the upper guide bearing is used as the y-axis coordinate value of the upper guide bearing in the three-dimensional coordinate system, and the relative distance between the upper guide bearing and the water guide bearing is used as the z-axis coordinate value of the upper guide bearing in the three-dimensional coordinate system, to obtain the three-dimensional coordinates of the upper guide bearing in the pumped storage unit at the historical moment;

[0208] The lateral displacement waveform data of the lower guide bearing in the pumped storage unit is used as the x-axis coordinate value of the lower guide bearing in the three-dimensional coordinate system, the longitudinal displacement waveform data of the lower guide bearing is used as the y-axis coordinate value of the lower guide bearing in the three-dimensional coordinate system, and the relative distance between the lower guide bearing and the water guide bearing is used as the z-axis coordinate value of the lower guide bearing in the three-dimensional coordinate system, to obtain the three-dimensional coordinates of the lower guide bearing in the pumped storage unit at the historical moment;

[0209] The lateral displacement waveform data of the water-guide bearing in the pumped-storage unit is used as the x-axis coordinate value of the water-guide bearing in the three-dimensional coordinate system, the longitudinal displacement waveform data of the water-guide bearing is used as the y-axis coordinate value of the water-guide bearing in the three-dimensional coordinate system, and 0 is used as the z-axis coordinate value of the water-guide bearing in the three-dimensional coordinate system, to obtain the three-dimensional coordinates of the water-guide bearing in the pumped-storage unit at the historical moment;

[0210] The curvature of the main shaft in the pumped storage unit at the historical moment is determined based on the three-dimensional coordinates of the upper, lower and water-guided bearings in the pumped storage unit at the historical moment.

[0211] Furthermore, the curvature of the main shaft in the water storage unit at the historical moment is as follows:

[0212]

[0213] In the above formula, S is the curvature of the main shaft in the water storage unit at the historical moment, s is the relative distance between the upper guide bearing and the water guide bearing, s′ is the vertical distance between the lower guide bearing and the straight line a, and the endpoints of the straight line a are the three-dimensional coordinates of the upper guide bearing and the three-dimensional coordinates of the water guide bearing, respectively.

[0214] Furthermore, the relative distance between the upper guide bearing and the water guide bearing is as follows:

[0215]

[0216] In the above formula, A 上 , B 上 and C 上 A is the three-dimensional coordinate x, y and z axis coordinate value of the upper guide bearing, 水 , B 水 and C 水 are the three-dimensional x, y and z axis coordinate values ​​of the water guide bearing.

[0217] Furthermore, the vertical distance between the lower guide bearing and the straight line a is as follows:

[0218]

[0219] In the above formula, P 上 is the three-dimensional coordinate of the upper guide bearing, P 下 is the three-dimensional coordinate of the lower guide bearing, P 水 is the three-dimensional coordinate of the water-guided bearing.

[0220] Preferably, the second analysis module is specifically used for:

[0221] When the curvature of the main shaft in the pumped-storage unit in the prediction period is a positive value, the main shaft in the pumped-storage unit is in a convex posture. When the curvature of the main shaft in the pumped-storage unit in the prediction period is a negative value, the main shaft in the pumped-storage unit is in a concave posture.

[0222] Furthermore, the initial convolutional neural network of the convolutional neural network model based on the attention mechanism and the long short-term memory network includes:

[0223] Layer 1: Fully connected layer, with 6 input features and 30 output features;

[0224] Layer 2: 2D convolutional layer, with 1 input channel, 64 output channels, stride 1, kernel size 3, and same padding;

[0225] Layer 3: ReLU activation function;

[0226] Layer 4: 2D max pooling layer with pooling size of 2, stride of 1, and same padding;

[0227] Layer 5: 2D dropout layer with dropout probability set to 0.3;

[0228] Layer 6: Long short-term memory network, input size is 64*30, hidden state size is 256, number of recurrent network layers is 1, time_major is set to False;

[0229] Layer 7: LSTM network, input size is 256, hidden state size is 128, number of recurrent layers is 1, time_major is set to False;

[0230] Layer 8: Long short-term memory network, input size is 128, hidden state size is 64, number of recurrent network layers is 1, time_major is set to False;

[0231] Layer 9: Fully connected layer, with 64 input features and 32 output features;

[0232] Layer 10: ReLU activation function;

[0233] Layer 11: Fully connected layer, with an input feature of 32 and an output feature of 1.

[0234] Furthermore, the attention mechanism of the initial convolutional neural network model based on the attention mechanism and the long short-term memory network includes: a channel attention model and a spatial attention model;

[0235] The channel attention model includes:

[0236] Layer 1: 2D adaptive average pooling layer, the output feature map size is 1;

[0237] Layer 2: 2D adaptive max pooling layer with an output feature map size of 1;

[0238] Layer 3: 2D convolutional layer;

[0239] Layer 4: activation function ReLU;

[0240] Layer 5: 2D convolutional layer;

[0241] Layer 6: activation function Sigmoid;

[0242] The spatial attention model includes:

[0243] Layer 1: 2D convolutional layer;

[0244] Layer 2: Activation function Sigmoid.

[0245] Furthermore, the channel attention model is placed in the third layer of the convolutional neural network: after the ReLU activation function, and the spatial attention model is placed after the channel attention model.

[0246] Example 3

[0247] Based on the same inventive concept, the present invention also provides a computer device, which includes a processor and a memory, wherein the memory is used to store a computer program, the computer program includes program instructions, and the processor is used to execute the program instructions stored in the computer storage medium. The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. It is the computing core and control core of the terminal, which is suitable for implementing one or more instructions, and is specifically suitable for loading and executing one or more instructions in the computer storage medium to implement the corresponding method flow or corresponding function, so as to implement the steps of a method for predicting the shaft system attitude of a pumped storage unit in the above embodiment.

[0248] Example 4

[0249] Based on the same inventive concept, the present invention also provides a storage medium, specifically a computer-readable storage medium (Memory), which is a memory device in a computer device for storing programs and data. It can be understood that the computer-readable storage medium here can include both built-in storage media in a computer device and, of course, extended storage media supported by the computer device. The computer-readable storage medium provides a storage space, which stores the operating system of the terminal. In addition, one or more instructions suitable for being loaded and executed by a processor are also stored in the storage space, and these instructions can be one or more computer programs (including program codes). It should be noted that the computer-readable storage medium here can be a high-speed RAM memory or a non-volatile memory, such as at least one disk memory. The processor can load and execute one or more instructions stored in the computer-readable storage medium to implement the steps of a method for predicting the shaft system attitude of a pumped storage unit in the above embodiment.

[0250] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0251] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0252] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0253] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0254] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for predicting the shaft system attitude of a pumped storage unit, characterized in that: The method comprises: The displacement waveform data of the bearing in the pumped storage unit in the current period is used as the input of the pre-trained convolutional neural network model based on the attention mechanism and the long short-term memory network, and the curvature of the main shaft in the pumped storage unit in the predicted period is obtained as the output of the pre-trained convolutional neural network model based on the attention mechanism and the long short-term memory network; The shaft system posture of the pumped-storage unit within the prediction period is analyzed based on the curvature of the main shaft in the pumped-storage unit within the prediction period.

2. The method according to claim 1, characterized in that The displacement waveform data of the bearing includes: the lateral displacement waveform data of the upper guide bearing, the longitudinal displacement waveform data of the upper guide bearing, the lateral displacement waveform data of the lower guide bearing, the longitudinal displacement waveform data of the lower guide bearing, the lateral displacement waveform data of the water guide bearing, and the longitudinal displacement waveform data of the water guide bearing.

3. The method according to claim 1, characterized in that The displacement waveform data of the bearing is obtained by an eddy current displacement sensor.

4. The method according to claim 2, characterized in that The training process of the pre-trained convolutional neural network model based on the attention mechanism and the long short-term memory network includes: Obtaining displacement waveform data of bearings in a pumped-storage unit in a historical period, and calculating standardized data corresponding to the displacement waveform data of the bearings in the historical period; determining the curvature of the main shaft in the hydroelectric storage unit during the historical period based on the standardized data; Constructing training data using displacement waveform data of bearings in the pumped storage unit in the historical period and curvature of the main shaft in the water storage unit in the historical period; The training data is used to train the initial convolutional neural network model based on the attention mechanism and the long short-term memory network to obtain the pre-trained convolutional neural network model based on the attention mechanism and the long short-term memory network.

5. The method according to claim 4, characterized in that The standardized data corresponding to the displacement waveform data of the bearing during the historical period are as follows: x s =x-μ In the above formula, x s is the standardized data corresponding to the displacement waveform data of the bearing in the historical period, x is the displacement waveform data of the bearing in the historical period, and μ is the mean value of the displacement waveform data of the bearing in the historical period.

6. The method according to claim 5, characterized in that Determining the curvature of the main shaft in the water storage unit in the historical period based on the standardized data includes: A three-dimensional coordinate system is constructed with the water guide bearing as the origin, the main axis direction as the z-axis, the upper guide bearing transverse direction, the lower guide bearing transverse direction or the water guide bearing transverse direction as the x-axis, and the upper guide bearing longitudinal direction, the lower guide bearing longitudinal direction or the water guide bearing longitudinal direction as the y-axis; The lateral displacement waveform data of the upper guide bearing in the pumped storage unit is used as the x-axis coordinate value of the upper guide bearing in the three-dimensional coordinate system, the longitudinal displacement waveform data of the upper guide bearing is used as the y-axis coordinate value of the upper guide bearing in the three-dimensional coordinate system, and the relative distance between the upper guide bearing and the water guide bearing is used as the z-axis coordinate value of the upper guide bearing in the three-dimensional coordinate system, to obtain the three-dimensional coordinates of the upper guide bearing in the pumped storage unit at the historical moment; The lateral displacement waveform data of the lower guide bearing in the pumped storage unit is used as the x-axis coordinate value of the lower guide bearing in the three-dimensional coordinate system, the longitudinal displacement waveform data of the lower guide bearing is used as the y-axis coordinate value of the lower guide bearing in the three-dimensional coordinate system, and the relative distance between the lower guide bearing and the water guide bearing is used as the z-axis coordinate value of the lower guide bearing in the three-dimensional coordinate system, to obtain the three-dimensional coordinates of the lower guide bearing in the pumped storage unit at the historical moment; The lateral displacement waveform data of the water-guide bearing in the pumped-storage unit is used as the x-axis coordinate value of the water-guide bearing in the three-dimensional coordinate system, the longitudinal displacement waveform data of the water-guide bearing is used as the y-axis coordinate value of the water-guide bearing in the three-dimensional coordinate system, and 0 is used as the z-axis coordinate value of the water-guide bearing in the three-dimensional coordinate system, to obtain the three-dimensional coordinates of the water-guide bearing in the pumped-storage unit at the historical moment; The curvature of the main shaft in the pumped storage unit at the historical moment is determined based on the three-dimensional coordinates of the upper, lower and water-guided bearings in the pumped storage unit at the historical moment.

7. The method according to claim 6, characterized in that The curvature of the main shaft in the water storage unit at the historical moment is as follows: In the above formula, S is the curvature of the main shaft in the water storage unit at the historical moment, s is the relative distance between the upper guide bearing and the water guide bearing, s′ is the vertical distance between the lower guide bearing and the straight line a, and the endpoints of the straight line a are the three-dimensional coordinates of the upper guide bearing and the three-dimensional coordinates of the water guide bearing, respectively.

8. The method according to claim 7, characterized in that The relative distance between the upper guide bearing and the water guide bearing is as follows: In the above formula, A 上 , B 上 and C 上 A is the three-dimensional coordinate x, y and z axis coordinate value of the upper guide bearing, 水 , B 水 and C 水 are the three-dimensional x, y and z axis coordinate values ​​of the water guide bearing.

9. The method according to claim 8, characterized in that The vertical distance between the lower guide bearing and the straight line a is as follows: In the above formula, P 上 is the three-dimensional coordinate of the upper guide bearing, P 下 is the three-dimensional coordinate of the lower guide bearing, P 水 is the three-dimensional coordinate of the water-guided bearing.

10. The method according to claim 1, characterized in that The method of analyzing the shaft system posture of the pumped-storage unit within the prediction period based on the curvature of the main shaft in the pumped-storage unit within the prediction period includes: When the curvature of the main shaft in the pumped-storage unit in the prediction period is a positive value, the main shaft in the pumped-storage unit is in a convex posture. When the curvature of the main shaft in the pumped-storage unit in the prediction period is a negative value, the main shaft in the pumped-storage unit is in a concave posture.

11. The method according to claim 4, characterized in that The initial convolutional neural network of the convolutional neural network model based on the attention mechanism and the long short-term memory network includes: Layer 1: Fully connected layer, with 6 input features and 30 output features; Layer 2: 2D convolutional layer, with 1 input channel, 64 output channels, stride 1, kernel size 3, and same padding. Layer 3: ReLU activation function; Layer 4: 2D max pooling layer with pooling size of 2, stride of 1, and same padding; Layer 5: 2D dropout layer with dropout probability set to 0.3; Layer 6: Long short-term memory network, input size is 64*30, hidden state size is 256, number of recurrent network layers is 1, time_major is set to False; Layer 7: LSTM network, input size is 256, hidden state size is 128, number of recurrent layers is 1, time_major is set to False; Layer 8: Long short-term memory network, input size is 128, hidden state size is 64, number of recurrent network layers is 1, time_major is set to False; Layer 9: Fully connected layer, with 64 input features and 32 output features; Layer 10: ReLU activation function; Layer 11: Fully connected layer, with an input feature of 32 and an output feature of 1.

12. The method according to claim 4, characterized in that The attention mechanism of the initial convolutional neural network model based on the attention mechanism and the long short-term memory network includes: a channel attention model and a spatial attention model; The channel attention model includes: Layer 1: 2D adaptive average pooling layer, the output feature map size is 1; Layer 2: 2D adaptive max pooling layer with an output feature map size of 1; Layer 3: 2D convolutional layer; Layer 4: activation function ReLU; Layer 5: 2D convolutional layer; Layer 6: activation function Sigmoid; The spatial attention model includes: Layer 1: 2D convolutional layer; Layer 2: Activation function Sigmoid.

13. The method according to claim 12, characterized in that The channel attention model is placed in the third layer of the convolutional neural network: after the ReLU activation function, and the spatial attention model is placed after the channel attention model.

14. A device for predicting the shaft system attitude of a pumped storage unit, characterized in that: The device comprises: The first analysis module is used to use the displacement waveform data of the bearing in the pumped-storage unit in the current period as the input of the pre-trained convolutional neural network model based on the attention mechanism and the long short-term memory network, and obtain the curvature of the main shaft in the pumped-storage unit in the predicted period output by the pre-trained convolutional neural network model based on the attention mechanism and the long short-term memory network; The second analysis module is used to analyze the shaft system posture of the pumped-storage unit within the prediction period based on the curvature of the main shaft in the pumped-storage unit within the prediction period.

15. The device according to claim 14, characterized in that The displacement waveform data of the bearing includes: the lateral displacement waveform data of the upper guide bearing, the longitudinal displacement waveform data of the upper guide bearing, the lateral displacement waveform data of the lower guide bearing, the longitudinal displacement waveform data of the lower guide bearing, the lateral displacement waveform data of the water guide bearing, and the longitudinal displacement waveform data of the water guide bearing.

16. The device according to claim 14, characterized in that The displacement waveform data of the bearing is obtained by an eddy current displacement sensor.

17. The device according to claim 15, characterized in that The training process of the pre-trained convolutional neural network model based on the attention mechanism and the long short-term memory network includes: Obtaining displacement waveform data of bearings in a pumped-storage unit in a historical period, and calculating standardized data corresponding to the displacement waveform data of the bearings in the historical period; determining the curvature of the main shaft in the hydroelectric storage unit during the historical period based on the standardized data; Constructing training data using displacement waveform data of bearings in the pumped storage unit in the historical period and curvature of the main shaft in the water storage unit in the historical period; The training data is used to train the initial convolutional neural network model based on the attention mechanism and the long short-term memory network to obtain the pre-trained convolutional neural network model based on the attention mechanism and the long short-term memory network.

18. The device according to claim 17, characterized in that The standardized data corresponding to the displacement waveform data of the bearing during the historical period are as follows: x s =x-μ In the above formula, x s is the standardized data corresponding to the displacement waveform data of the bearing in the historical period, x is the displacement waveform data of the bearing in the historical period, and μ is the mean value of the displacement waveform data of the bearing in the historical period.

19. The device according to claim 18, characterized in that Determining the curvature of the main shaft in the water storage unit in the historical period based on the standardized data includes: A three-dimensional coordinate system is constructed with the water guide bearing as the origin, the main axis direction as the z-axis, the upper guide bearing transverse direction, the lower guide bearing transverse direction or the water guide bearing transverse direction as the x-axis, and the upper guide bearing longitudinal direction, the lower guide bearing longitudinal direction or the water guide bearing longitudinal direction as the y-axis; The lateral displacement waveform data of the upper guide bearing in the pumped storage unit is used as the x-axis coordinate value of the upper guide bearing in the three-dimensional coordinate system, the longitudinal displacement waveform data of the upper guide bearing is used as the y-axis coordinate value of the upper guide bearing in the three-dimensional coordinate system, and the relative distance between the upper guide bearing and the water guide bearing is used as the z-axis coordinate value of the upper guide bearing in the three-dimensional coordinate system, to obtain the three-dimensional coordinates of the upper guide bearing in the pumped storage unit at the historical moment; The lateral displacement waveform data of the lower guide bearing in the pumped storage unit is used as the x-axis coordinate value of the lower guide bearing in the three-dimensional coordinate system, the longitudinal displacement waveform data of the lower guide bearing is used as the y-axis coordinate value of the lower guide bearing in the three-dimensional coordinate system, and the relative distance between the lower guide bearing and the water guide bearing is used as the z-axis coordinate value of the lower guide bearing in the three-dimensional coordinate system, to obtain the three-dimensional coordinates of the lower guide bearing in the pumped storage unit at the historical moment; The lateral displacement waveform data of the water-guide bearing in the pumped-storage unit is used as the x-axis coordinate value of the water-guide bearing in the three-dimensional coordinate system, the longitudinal displacement waveform data of the water-guide bearing is used as the y-axis coordinate value of the water-guide bearing in the three-dimensional coordinate system, and 0 is used as the z-axis coordinate value of the water-guide bearing in the three-dimensional coordinate system, to obtain the three-dimensional coordinates of the water-guide bearing in the pumped-storage unit at the historical moment; The curvature of the main shaft in the pumped storage unit at the historical moment is determined based on the three-dimensional coordinates of the upper, lower and water-guided bearings in the pumped storage unit at the historical moment.

20. The device according to claim 19, characterized in that The curvature of the main shaft in the water storage unit at the historical moment is as follows: In the above formula, S is the curvature of the main shaft in the water storage unit at the historical moment, s is the relative distance between the upper guide bearing and the water guide bearing, s′ is the vertical distance between the lower guide bearing and the straight line a, and the endpoints of the straight line a are the three-dimensional coordinates of the upper guide bearing and the three-dimensional coordinates of the water guide bearing, respectively.

21. The device according to claim 20, characterized in that The relative distance between the upper guide bearing and the water guide bearing is as follows: In the above formula, A 上 , B 上 and C 上 is the three-dimensional coordinate x, y and z axis coordinate value of the upper guide bearing, a 水 , B 水 and C 水 are the three-dimensional x, y and z axis coordinate values ​​of the water guide bearing.

22. The device according to claim 21, characterized in that The vertical distance between the lower guide bearing and the straight line a is as follows: In the above formula, P 上 is the three-dimensional coordinate of the upper guide bearing, P 下 is the three-dimensional coordinate of the lower guide bearing, P 水 is the three-dimensional coordinate of the water-guided bearing.

23. The device according to claim 14, characterized in that The second analysis module is specifically used for: When the curvature of the main shaft in the pumped-storage unit in the prediction period is a positive value, the main shaft in the pumped-storage unit is in a convex posture. When the curvature of the main shaft in the pumped-storage unit in the prediction period is a negative value, the main shaft in the pumped-storage unit is in a concave posture.

24. The device according to claim 17, characterized in that The initial convolutional neural network of the convolutional neural network model based on the attention mechanism and the long short-term memory network includes: Layer 1: Fully connected layer, with 6 input features and 30 output features; Layer 2: 2D convolutional layer, with 1 input channel, 64 output channels, stride 1, kernel size 3, and same padding; Layer 3: ReLU activation function; Layer 4: 2D max pooling layer with pooling size of 2, stride of 1, and same padding; Layer 5: 2D dropout layer with dropout probability set to 0.3; Layer 6: Long short-term memory network, input size is 64*30, hidden state size is 256, number of recurrent network layers is 1, time_major is set to False; Layer 7: LSTM network, input size is 256, hidden state size is 128, number of recurrent layers is 1, time_major is set to False; Layer 8: Long short-term memory network, input size is 128, hidden state size is 64, number of recurrent network layers is 1, time_major is set to False; Layer 9: Fully connected layer, with 64 input features and 32 output features; Layer 10: ReLU activation function; Layer 11: Fully connected layer, with an input feature of 32 and an output feature of 1.

25. The device according to claim 17, characterized in that The attention mechanism of the initial convolutional neural network model based on the attention mechanism and the long short-term memory network includes: a channel attention model and a spatial attention model; The channel attention model includes: Layer 1: 2D adaptive average pooling layer, the output feature map size is 1; Layer 2: 2D adaptive max pooling layer with an output feature map size of 1; Layer 3: 2D convolutional layer; Layer 4: activation function ReLU; Layer 5: 2D convolutional layer; Layer 6: activation function Sigmoid; The spatial attention model includes: Layer 1: 2D convolutional layer; Layer 2: Activation function Sigmoid.

26. The device according to claim 25, characterized in that The channel attention model is placed in the third layer of the convolutional neural network: after the ReLU activation function, and the spatial attention model is placed after the channel attention model.

27. A computer device, characterized in that: include: one or more processors; The processor is configured to execute one or more programs; When the one or more programs are executed by the one or more processors, the method for predicting the shaft system posture of a pumped-storage unit as described in any one of claims 1 to 13 is implemented.

28. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the computer program is executed, the method for predicting the shaft system posture of a pumped storage unit as described in any one of claims 1 to 13 is implemented.

Images