Temperature control method, device and equipment for water guide bearing bush of water-turbine generator set and medium

Abstract

The invention discloses a water-turbine generator set water guide bearing bush temperature control method, device and equipment and a medium, and the method comprises the steps: extracting local features from original time sequence data through a convolutional neural network, capturing a short-term dependency relationship in temperature change, processing the local features through a long and short-term memory network, capturing a long-term dependency relationship in temperature data, and obtaining a temperature control result; meanwhile, channel weights of local features are adaptively recalibrated in combination with a channel attention network, so that the feature expression ability of the model is further improved, and the optimization problem is designed for a water guide bearing bush temperature prediction model and a data-driven model prediction control and multi-model switching control strategy. Rolling optimal control is carried out based on a gradient descent method, the temperature control precision is improved, and then the safety of the hydropower station is improved.

Description

Technical Field

[0001] The invention relates to temperature control technology, and in particular to a method, device, equipment and medium for controlling the temperature of a water guide bearing of a hydro-turbine generator set. Background Art

[0002] In the process of hydroelectric power generation, the turbine is an important key part. Among them, the water guide bearing is one of the key components of the turbine, and its stability directly affects the performance and life of the turbine.

[0003] The water guide bearing temperature of the turbine is one of the important parameters to measure the operating status and health of the turbine. Excessive water guide bearing temperature may lead to lubricant deterioration, material deformation and even equipment failure, thus affecting the normal operation and power generation efficiency of the hydropower station. Therefore, accurate control of the water guide bearing temperature of the turbine is of great significance to ensure the safe operation of the hydropower station and improve the economic benefits.

[0004] Traditional water-guided bearing temperature control methods mainly rely on statistical models and empirical formulas. These methods often perform poorly when dealing with nonlinear and non-stationary time series data and have limited control accuracy. Summary of the invention

[0005] The present invention provides a method, device, equipment and medium for controlling the temperature of a water-turbine generator water guide bearing, so as to improve the temperature control accuracy and the safety of a hydropower station.

[0006] In a first aspect, the present invention provides a method for controlling the temperature of a water guide bearing of a hydro-turbine generator set, comprising:

[0007] Collect historical temperature data of water guide bearings of hydro-turbine generator sets;

[0008] Inputting the historical temperature data into a water-guide bearing temperature prediction model for processing, and predicting the predicted temperature of the water-guide bearing at the next control moment, wherein the water-guide bearing temperature prediction model integrates a convolutional neural network, a channel attention network, and a long short-term memory network;

[0009] Based on the predicted temperature, model predictive control is used to optimize control parameters of the hydro-generator set.

[0010] Optionally, before inputting the historical temperature data into the water guide bearing temperature prediction model for processing, the method further includes:

[0011] Performing data cleaning on the historical temperature data to remove abnormal data;

[0012] Resample the cleaned historical temperature data to a fixed time interval;

[0013] Perform linear interpolation on the resampled historical temperature data to align the data.

[0014] Optionally, the historical temperature data is input into a water-guide bearing temperature prediction model for processing to predict the predicted temperature of the water-guide bearing at the next control moment, including:

[0015] Inputting the historical temperature data into the convolutional neural network for processing to obtain local features of the historical temperature data;

[0016] Inputting the local feature into the channel attention network for processing to obtain a channel attention weight vector, wherein the channel attention weight vector includes the channel attention weights of each channel of the local feature;

[0017] Multiplying the local feature by the channel attention weight vector to obtain a channel attention feature;

[0018] Inputting the channel attention feature into the long short-term memory network for processing to obtain the time series feature of the historical temperature data;

[0019] The predicted temperature of the water guide bearing at the next control moment is predicted based on the time series characteristics.

[0020] Optionally, based on the predicted temperature, using model predictive control to optimize control parameters of the hydro-generator set includes:

[0021] Taking the unit load and the coolant level as inputs of the model predictive control, calculating the cost function value of the model predictive control;

[0022] Determine whether the cost function value is greater than a threshold;

[0023] If so, the unit load and the coolant level are updated based on the predicted temperature, the parameters of the model predictive control are adaptively learned, and the step of calculating the cost function value of the model predictive control by taking the unit load and the coolant level as the input of the model predictive control is returned to be executed until the cost function value is less than or equal to the threshold value, or the number of iterations reaches a preset number;

[0024] If not, the control parameters of the hydro-generator set are output.

[0025] Optionally, updating the unit load and the coolant level based on the predicted temperature includes:

[0026] Calculate the Jacobian matrix of the water-guide bearing temperature prediction model based on the predicted temperature;

[0027] Based on the Jacobian matrix, the unit load and the coolant level are updated using a gradient descent method.

[0028] Optionally, based on the predicted temperature, the control parameters of the hydro-generator set are optimized by using model predictive control to convert the optimization problem into an objective function:

[0029]

[0030] |Δu(t)|≤Δu max ;

[0031] u min ≤u(t)≤u max ;

[0032]

[0033] in, is the objective function, a and b are weight parameters, and the historical input is U p (t)=[u(t),u(t-1),...,u(tl) u )], the historical output is Y p =[y(t),y(t-1),...,y(tl y )], and R(t)=[r(t+1), r(t+2),..., r(t+T p )] is the reference output, is the predicted output, U(t) = [u(t), u(t+1), ..., u(t+T c -1)] is the optimal control input, ΔU(t) = [Δu(t), Δu(t+1), ..., Δu(t+T c -1)] is the incremental control movement.

[0034] Optionally, calculating the Jacobian matrix of the water guide bearing temperature prediction model based on the predicted temperature includes:

[0035] The derivative of the output of the water-guided bearing temperature prediction model is calculated as the Jacobian matrix of the water-guided bearing temperature prediction model.

[0036] In a second aspect, the present invention further provides a water guide bearing temperature control device for a hydro-generator set, comprising:

[0037] A data acquisition module is used to collect historical temperature data of the water guide bearing of the hydro-turbine generator set;

[0038] A temperature prediction module, used for inputting the historical temperature data into a water-guided bearing temperature prediction model for processing, and predicting the predicted temperature of the water-guided bearing at the next control moment, wherein the water-guided bearing temperature prediction model integrates a convolutional neural network, a channel attention network, and a long short-term memory network;

[0039] A parameter optimization module is used to optimize the control parameters of the hydro-generator set by adopting model predictive control based on the predicted temperature.

[0040] In a third aspect, the present invention further provides an electronic device, comprising:

[0041] one or more processors;

[0042] A storage device for storing one or more programs;

[0043] When the one or more programs are executed by the one or more processors, the one or more processors implement the water guide bearing temperature control method of the hydro-turbine generator set as provided in the first aspect of the present invention.

[0044] In a fourth aspect, the present invention further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for controlling the temperature of a water guide bearing of a hydro-turbine generator set as provided in the first aspect of the present invention.

[0045] The water guide bearing temperature control method of a hydro-turbine generator set provided by the present invention utilizes a convolutional neural network to extract local features from original time series data to capture short-term dependencies in temperature changes, utilizes a long short-term memory network to process local features to capture long-term dependencies in temperature data, and combines a channel attention network to adaptively recalibrate the channel weights of local features, thereby further improving the feature expression capability of the model. For the water guide bearing temperature prediction model, data-driven model predictive control and multi-model switching control strategies are used to design optimization problems, and rolling optimal control is performed based on the gradient descent method to improve temperature control accuracy, thereby improving the safety of the hydropower station.

[0046] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present invention, nor are they intended to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0048] Figure 1 A flow chart of a method for controlling the temperature of a water guide bearing of a hydro-turbine generator set provided by the present invention;

[0049] Figure 2A schematic diagram of the structure of a water-guided bearing temperature prediction model provided by the present invention;

[0050] Figure 3 A schematic diagram of the structure of a convolutional neural network provided by the present invention;

[0051] Figure 4 A schematic diagram of the processing process of a channel attention network provided by the present invention;

[0052] Figure 5 A schematic diagram of the structure of a long short-term memory network provided by the present invention;

[0053] Figure 6 A schematic diagram of the framework structure of a multi-model MPC provided by the present invention;

[0054] Figure 7 A schematic diagram of the structure of a water guide bearing temperature control device for a hydro-generator set provided by the present invention;

[0055] Figure 8 The present invention provides a schematic structural diagram of an electronic device.

[0056] The above drawings have shown clear embodiments of the present application, which will be described in more detail later. These drawings and text descriptions are not intended to limit the scope of the present application in any way, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0057] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme 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 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 should fall within the scope of protection of the present invention.

[0058] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0059] Figure 1 This is a flowchart of a method for controlling the temperature of the water guide bearing of a hydro-generator set provided by the present invention. This embodiment is applicable to the situation of processing non-linear and non-stationary time series data and improving the temperature control accuracy of the water guide bearing. This method can be executed by the water guide bearing temperature control device provided by the present invention, and this device can be implemented in the form of software and / or hardware, and is usually configured in an electronic device, such as Figure 1 As shown, the method for controlling the temperature of the water guide bearing of a hydro-generator set includes the following steps:

[0060] S101. Collect historical temperature data of the water guide bearing of the hydro-generator set.

[0061] In the embodiment of the present invention, taking a pumped-storage power station as the research object, the historical temperature data includes the temperature of the water guide bearing of the hydro-generator from May 2023 to May 2024, a total of 24,537 pieces. Due to reasons such as data loss, sensor failure, and input error, there are certain abnormal data in the historical temperature data, which will lead to many problems such as unstable model training, inaccurate prediction results, and incorrect prediction result trends. Therefore, it is crucial to clean the abnormal data. Use statistics such as mean and standard deviation to identify and process abnormal data, and regard the data exceeding twice the mean as abnormal data and clean it to increase the stability and reliability of the training data.

[0062] Due to reasons such as different data collection frequencies and data loss of the water guide bearing temperature data of the hydro-generator, the data at different time points are not synchronized. Therefore, data alignment is required to ensure the consistency of the data in the time dimension, so that the model can effectively use this data for training and prediction. First, resample the data to ensure that the data is unified to a fixed time interval to ensure the consistency of the data in the time dimension, making the subsequent interpolation and model training more accurate and effective. Through the observation and analysis of the training data, the present invention adopts the method of downsampling to fix the data with a short and inconsistent time interval to the sampling time T = 1h.

[0063] Considering that the change of the bearing temperature is relatively stable, a linear interpolation method is adopted for data alignment and data resampling. Assume that there are known data points y1 and y2 at time points t1 and t2, and there is a missing data point y at time t, where t1 < t < t2. Then the formula for linear interpolation is:

[0064]

[0065] After completing the data alignment, in order to improve the accuracy and robustness of univariate time series prediction, the present invention sets the input data to the five observations before the current moment, and the output is the observation value of the next moment. The data processing method is shown in the following two formulas:

[0066] X t =[y t-5 ,y t-4 ,y t-3 ,y t-2 ,y t-1 ]

[0067] Y t =y t

[0068] Among them, X t is the input data of the network model, Y t is the output of the network model, y t is the bearing temperature data at the next moment, [y t-5 ,y t-4 ,y t-3 ,y t-2 ,y t-1 ] represents the historical data of bearing temperature in the five moments before the current moment.

[0069] S102, inputting the historical temperature data into the water-guided bearing temperature prediction model for processing, and predicting the predicted temperature of the water-guided bearing at the next control moment.

[0070] Figure 2 A schematic diagram of a water-guided bearing temperature prediction model provided by the present invention is shown in FIG. Figure 2 As shown, in an embodiment of the present invention, the water-guide bearing temperature prediction model integrates convolutional neural networks (CNN), squeeze-and-excitation attention networks (SEA) and long short-term memory networks (LSTM).

[0071] Specifically, Figure 2As shown, the historical temperature data is input into the convolutional neural network for processing to obtain the local features of the historical temperature data, and the local features are input into the channel attention network (SE-Attention in the figure) for processing to obtain the channel attention weight vector. The channel attention weight vector includes the channel attention weights of each channel of the local features. The local features are multiplied by the channel attention weight vector to obtain the channel attention features, and the channel attention features are input into the long short-term memory network for processing to obtain the time series features of the historical temperature data. The predicted temperature of the water guide bearing at the next control moment is predicted based on the time series features.

[0072] Figure 3 A schematic diagram of the structure of a convolutional neural network provided by the present invention is shown in FIG. Figure 3 As shown in the figure, convolutional neural network is a widely used branch in the field of deep learning. It can efficiently extract features from data and has advantages such as invariance and noise resistance. The basic structure of convolutional neural network includes input layer, convolution layer, pooling layer, fully connected layer and output layer, among which the convolution layer is responsible for convolution operation on data and extracting potential features, while the pooling layer is used to downsample and compress network parameters. By alternating between convolutional layer and pooling layer, convolutional neural network can maximize the extraction of potential features of input data while reducing the errors caused by artificial feature extraction.

[0073] When dealing with the time series problem of the temperature of the bearing of a hydro-turbine generator, the convolutional neural network has the following advantages. First, the convolutional neural network can enhance the ability to capture local features and effectively extract information about sudden events in the local time series. Secondly, the convolutional neural network has high parameter efficiency. Since its convolution kernel remains unchanged during calculation and has the characteristics of shared weights, coupled with the characteristics of local connection, the number of parameters of the model is reduced and the training efficiency is higher. In addition, the convolutional neural network has parallel computing capabilities and can extract features from multiple local areas at the same time. The calculations at different time steps are independent and do not need to be performed in sequence, thereby speeding up the training speed. Finally, the convolution operation of the convolutional neural network can smooth out these interferences when processing interference noise or outliers in one-dimensional data, while retaining effective information, thereby enhancing the robustness of the model.

[0074] Figure 4 A schematic diagram of the processing process of a channel attention network provided by the present invention is as follows: Figure 4 As shown in Figure 1, the channel attention network is a channel type attention network that introduces an attention mechanism on the channel dimension. Figure 4Different colors represent different weights, so that each feature channel has different importance, so that the convolutional neural network can selectively focus on the channels with larger weights. This mechanism effectively solves the problem of feature loss caused by different proportions of different channels during convolution and pooling. The channel attention network is mainly divided into three parts: squeezing, excitation, and attention. First, through the squeezing operation, feature compression is performed from the spatial dimension, and the local features of size H×W×C are converted into feature vectors of size 1×1×C, so as to obtain the global receptive field. The number of channels of this vector matches the number of channels of the input local features, representing the global response distribution on the feature channel. This step is achieved through global average pooling. Next is the excitation operation, which generates weights for each feature channel by introducing the parameter W. W is a learnable multi-layer perceptron, which undergoes dimensionality reduction to reduce the number of parameters, and obtains normalized weights through the Sigmoid function to explicitly model the correlation between feature channels. Finally, through the attention operation, the weight of the excitation output is regarded as the importance of each feature channel, and is weighted to the original feature (i.e., local feature) by multiplying the channel width, completing the recalibration in the channel dimension.

[0075] Figure 5 A schematic diagram of the structure of a long short-term memory network provided by the present invention is shown in FIG. Figure 5 As shown in the figure, the long short-term memory network is an improved model based on the recurrent neural network (RNN). It adopts a three-layer gated design to better solve the problems of RNN gradient disappearance and gradient explosion. The long short-term memory network makes appropriate adjustments to the calculation of the hidden state and adds the cell state C as a memory unit to associate information with a longer time span. It will also forget part of the memory of previous information and has excellent predictive ability. The gated structure and memory unit of the long short-term memory network can help it effectively process long time series to achieve long-term preservation of historical information of the time series. The calculation formula of the long short-term memory network is:

[0076] f t =σ(W fx x t +W fh h t-1 +b f )

[0077] i t =σ(W ix x t +W ih h t-1 +b i )

[0078] g t =tanh(W gx xt +W gh h t-1 +b g )

[0079] O t =σ(W ox x t +W oh h t-1 +b o )

[0080] C t =g t i t +C t-1 f t

[0081] h t =tanh(C t ) t

[0082] Among them, W is the weight; b is the bias; σ is the Sigmoid function; C t-1 is the unit state at the previous moment; h t-1 is the intermediate state of the previous moment; x t is the input at the current moment; f t is the output of the forget gate; i t and O t are the outputs of the input and output gates respectively; g t is the output of the input node; C t is the unit state; h t It is an intermediate state.

[0083] The three gate control units are the forget gate, input gate and output gate. The forget gate determines which past information needs to be retained to prevent useless information from being passed on. t-1 and x t After weighted calculation, Sigmoid is used to perform nonlinear mapping to a value of 0 to 1 to get f t , used for the cell state C of the previous time step t-1 A certain degree of memory and forgetting, in which W fh The weights are not shared. The function of the input gate is to read the data and decide whether to update it to the memory cell. The Tanh layer generates candidate new information to be input, and the σ layer determines the degree of update of the cell state by generating a value of 0 to 1. The function of the output gate is to pass the processed data to the next moment, and generate a value of 0 to 1 through the Sigmoid function to determine how much the cell state is filtered before output. The cell state is first activated by the Tanh function, and after activation, it is connected with the O layer that controls which parts of the output. tMultiply them together and finally get the hidden state h that is input to the next time step t , and derive h t And after adding certain weights, we get the output of this step.

[0084] It can be seen that the activation functions of the input and output gates all use the Tanh function. This is because the second-order derivative of Tanh is not equal to zero in a large interval, which can prevent the gradient from disappearing. Its convergence speed is usually faster, the derivative is simple, and it takes up less computing resources. Its monotonous characteristics make it easy to transform the problem into a convex optimization problem. In addition, the core of the long short-term memory network structure is the number of cell states. The cell state is like a conveyor belt that runs through the entire process. The whole process does not change much, while the hidden state changes greatly. Therefore, the cell state is mainly used to memorize previous information. The cell update process is: the cell state from the previous time step first passes through the f generated by the forget gate t Filter the information to be left, and then enter the information through the i t Control the new information to be injected, and then C t Output to the next time step, and after activation for output. In general, the cell state mainly undergoes the process of screening and summation, so the overall change is not significant.

[0085] The LSTM network has several significant advantages: First, it performs better in modeling and feature extraction of long-term dependency data. Second, the LSTM network can adapt to the input of variable-length sequences and flexibly process data of different lengths. In addition, the LSTM network solves the gradient explosion and gradient vanishing problems to a certain extent. In the application of turbine generator bearing temperature, the LSTM network can effectively capture and memorize data dependencies over a long span, and is suitable for processing long-term trends and periodic changes in time series data. Convolutional neural networks are good at extracting local features from data, and can capture local patterns and trends in input data through convolution operations. Before introducing the convolutional neural network into the LSTM network model, the convolutional neural network is first used to extract features, and then the LSTM network is used for time series modeling, which can improve the accuracy of the prediction. This method can better capture the complex patterns and changing trends in the turbine generator bearing temperature data.

[0086] In order to obtain the best prediction model parameters, the present invention designs a control variable method to gradually tune the model, and gradually increases the number of layers of the long short-term memory network to test the impact of the increase in model depth on the prediction effect. First, the number of convolutional neural network layers is fixed, and the impact of long short-term memory networks with different numbers of layers and neurons on the prediction results is tested to obtain the best long short-term memory network parameters; then the long short-term memory network structure is fixed, and the convolutional neural network design is adjusted to obtain the best parameters.

[0087] S103. Based on the predicted temperature, the control parameters of the hydro-generator set are optimized using model predictive control.

[0088] Model Predictive Control (MPC) is a model-based control strategy that is widely used in the control and optimization of industrial processes. Although MPC has been proven to be effective in industrial processes, it still has some limitations. As industrial processes are becoming large-scale, complex, and highly coupled, different system parameters or structures can lead to different operating modes and bring about multi-mode control problems. For each mode, there are different operating parameters. If the control strategy does not match the corresponding mode, it will reduce the stability of the system and even cause serious production accidents and losses. Some improved MPC methods designed to handle multi-mode processes are called multi-model MPC or switching MPC. These methods establish multiple models for each operating mode and design model switching strategies. Figure 6 A schematic diagram of the framework structure of a multi-model MPC provided by the present invention is shown in FIG. Figure 6 As shown in Figure 1, the switching MPC of multi-mode control usually consists of three parts: prediction model group, controller group and decision unit. Its working process is as follows: First, a series of prediction models need to be established to track the behavior of the controlled system in different operating modes. Then, a switching strategy is designed in the decision unit to distinguish different operating modes. On this basis, the optimal controller can be obtained by solving the problem to ensure the stability of the whole process in each operating mode.

[0089] 1. Take the unit load and coolant level as the input of model predictive control and calculate the cost function value of model predictive control. In addition, set the reference trajectory, optimization termination condition threshold and number of iterations.

[0090] 2. Determine whether the cost function value is greater than the threshold.

[0091] 3. If so, update the unit load and coolant level based on the predicted temperature, adaptively learn the parameters of the model predictive control, and return to the step of calculating the cost function value of the model predictive control with the unit load and coolant level as the input of the model predictive control until the cost function value is less than or equal to the threshold, or the number of iterations reaches the preset number.

[0092] 4. If not, output the control parameters of the hydro-generator set.

[0093] Specifically, the goal of MPC is to find a suitable control signal u(t) through online optimization so that the system output tracks the reference trajectory r(t) as closely as possible. p and T c , the optimization problem can be expressed as the optimization problem of the objective function:

[0094]

[0095] |Δu(t)|≤Δu max

[0096] u min ≤u(t)≤u max

[0097]

[0098] in, is the objective function, a and b are weight parameters, and the historical input is U p (t)=[u(t),u(t-1),...,u(tl) u )], the historical output is Y p =[y(t),y(t-1),...,y(tl y )], and R(t)=[r(t+1), r(t+2),..., r(t+T p )] is the reference output, is the predicted output, U(t) = [u(t), u(t+1), ..., u(t+T c -1)] is the optimal control input, ΔU(t) = [Δu(t), Δu(t+1), ..., Δu(t+T c -1)] is the incremental control movement.

[0099] g(·) represents the prediction model based on SE-CNN-LSTM network, which can be expressed as:

[0100]

[0101] The present invention selects the unit load and the coolant level as the input of the MPC controller for the following reasons: First, during the operation of the generator, if the unit load is too heavy, the generator will generate too much heat, which will cause the bearing temperature to be too high. At this time, the unit load needs to be adjusted to reduce the overload phenomenon, thereby reducing the bearing temperature. Secondly, the generator needs to be cooled during operation. If the coolant level is insufficient, it may lead to insufficient cooling of the bearing, which will lead to excessive bearing temperature. At this time, it is necessary to check the coolant level, add coolant in time, and maintain the cooling system.

[0102] In the embodiment of the present invention, since the proposed prediction method is based on deep learning, its mathematical expression cannot be directly expressed, which brings challenges to solving the optimization problem. In order to reduce the number of iterations and quickly find the optimal result, an adaptive gradient descent method is selected to solve the optimization problem:

[0103] Uk+1 (t) = U k (t)+ΔU k (t)

[0104]

[0105] Where η1>0 is the learning rate and k is the number of iterations. The derivative of the objective function J(t) can be rewritten as:

[0106]

[0107] therefore,

[0108]

[0109] It can be found that the key to dealing with optimization problems is to calculate the derivative of the prediction model output, that is, the Jacobian matrix Because the activation functions used in the water-guide bearing temperature prediction model are continuously differentiable Sigmoid functions and Tanh functions, it is feasible to calculate the Jacobian matrix of the water-guide bearing temperature prediction model. In addition, numerical differentiation is used as an alternative method to ensure that every element in the Jacobian matrix is ​​available.

[0110] For the gradient descent method, constraints can be managed by the projected gradient method, which projects the optimization variables onto the allowed hyperspace. Since the constraints in the proposed optimization problem are box constraints (constant or linear constraints), we can obtain:

[0111] U k+1 (t) = P1(U k (t)+ΔU k (t))

[0112]

[0113] in is the vector u k (t), Δu k The projection function of an element of (t) can be defined as

[0114]

[0115] where u max is u k (t) is the upper bound constraint, u min is u k (t) is the lower bound constraint, Δu max is the incremental control input Δu k(t). When it comes to system state constraints, how to deal with control input constraints, add prediction methods to the prediction model to ensure that the constraints are satisfied. In order to speed up the optimization process, we propose an improved adaptive gradient descent method that uses the natural logarithm to decay the learning rate, thereby improving the convergence speed of the method:

[0116]

[0117] Among them, α is the decay rate of each iteration, and k is the number of iterations.

[0118] By solving the optimization problem, the optimal control input sequence can be obtained, and then the first element of U(t) is applied as the control signal in the system to ensure that the system output can accurately track the reference trajectory under different launch modes.

[0119] The water guide bearing temperature control method of a hydro-turbine generator set provided by the present invention utilizes a convolutional neural network to extract local features from original time series data to capture short-term dependencies in temperature changes, utilizes a long short-term memory network to process local features to capture long-term dependencies in temperature data, and combines a channel attention network to adaptively recalibrate the channel weights of local features, thereby further improving the feature expression capability of the model. For the water guide bearing temperature prediction model, data-driven model predictive control and multi-model switching control strategies are used to design optimization problems, and rolling optimal control is performed based on the gradient descent method to improve temperature control accuracy, thereby improving the safety of the hydropower station.

[0120] Figure 7 The schematic diagram of the structure of a water guide bearing temperature control device for a water turbine generator set provided by the present invention is as follows: Figure 7 As shown, the water guide bearing temperature control device of the hydro-generator set includes:

[0121] The data acquisition module 201 is used to collect the historical temperature data of the water guide bearing of the hydro-generator set;

[0122] The temperature prediction module 202 is used to input the historical temperature data into the water-guide bearing temperature prediction model for processing, and predict the predicted temperature of the water-guide bearing at the next control moment, wherein the water-guide bearing temperature prediction model integrates a convolutional neural network, a channel attention network, and a long short-term memory network;

[0123] The parameter optimization module 203 is used to optimize the control parameters of the hydro-generator set by using model predictive control based on the predicted temperature.

[0124] In some embodiments of the present invention, the water guide bearing temperature control device of the hydro-generator set further comprises:

[0125] A data cleaning module, used for cleaning the historical temperature data and eliminating abnormal data before inputting the historical temperature data into the water-guide bearing temperature prediction model for processing;

[0126] The resampling module is used to resample the cleaned historical temperature data to a fixed time interval;

[0127] The data alignment module is used to perform linear interpolation and data alignment on the resampled historical temperature data.

[0128] In some embodiments of the present invention, the temperature prediction module 202 includes:

[0129] A local feature extraction submodule, used for inputting the historical temperature data into the convolutional neural network for processing to obtain local features of the historical temperature data;

[0130] An attention processing submodule, used for inputting the local features into the channel attention network for processing to obtain a channel attention weight vector, wherein the channel attention weight vector includes the channel attention weights of each channel of the local features;

[0131] A product submodule, used for multiplying the local feature with the channel attention weight vector to obtain a channel attention feature;

[0132] A time series feature extraction submodule, used for inputting the channel attention feature into the long short-term memory network for processing to obtain the time series feature of the historical temperature data;

[0133] The temperature prediction submodule is used to predict the predicted temperature of the water guide bearing at the next control moment based on the time series characteristics.

[0134] In some embodiments of the present invention, the parameter optimization module 203 includes:

[0135] A cost function calculation submodule, used to calculate the cost function value of the model predictive control by taking the unit load and the coolant level as inputs of the model predictive control;

[0136] A judgment submodule, used to judge whether the cost function value is greater than a threshold;

[0137] Returning to the execution submodule, when the cost function value is greater than a threshold, updating the unit load and the coolant level based on the predicted temperature, adaptively learning the parameters of the model predictive control, and returning to the execution step of taking the unit load and the coolant level as inputs of the model predictive control and calculating the cost function value of the model predictive control, until the cost function value is less than or equal to the threshold, or the number of iterations reaches a preset number;

[0138] The parameter output submodule is used to output the control parameters of the hydro-generator set when the cost function value is less than or equal to a threshold value.

[0139] In some embodiments of the present invention, the return execution submodule includes:

[0140] A Jacobi matrix calculation unit, used for calculating the Jacobi matrix of the water guide bearing temperature prediction model based on the predicted temperature;

[0141] A parameter updating unit is used to update the unit load and the coolant level based on the Jacobi matrix by using a gradient descent method.

[0142] In some embodiments of the present invention, based on the predicted temperature, the control parameters of the hydro-generator set are optimized by using model predictive control, which is converted into an optimization problem of an objective function:

[0143]

[0144] |Δu(t)|≤Δu max ;

[0145] u min ≤u(t)≤u max ;

[0146]

[0147] in, is the objective function, a and b are weight parameters, and the historical input is U p (t)=[u(t),u(t-1),...,u(tl) u )], the historical output is Y p =[y(t),y(t-1),...,y(tl y )], and R(t)=[r(t+1), r(t+2),..., r(t+T p )] is the reference output, is the predicted output, U(t) = [u(t), u(t+1), ..., u(t+T c -1)] is the optimal control input, ΔU(t) = [Δu(t), Δu(t+1), ..., Δu(t+T c -1)] is the incremental control movement.

[0148] In some embodiments of the present invention, the Jacobi matrix calculation unit includes:

[0149] The calculation subunit is used to calculate the derivative of the output of the water-guided bearing temperature prediction model as the Jacobian matrix of the water-guided bearing temperature prediction model.

[0150] The above-mentioned water-turbine generator set water-guide bearing temperature control device can execute the water-turbine generator set water-guide bearing temperature control method provided by the aforementioned embodiment of the present invention, and has corresponding functional modules and beneficial effects for executing the water-turbine generator set water-guide bearing temperature control method.

[0151] Figure 8 A schematic diagram of the structure of an electronic device provided by the present invention. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or required herein.

[0152] like Figure 8 As shown, the electronic device includes at least one processor 11, and a memory connected to the at least one processor 11, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc., wherein the memory stores a computer program that can be executed by at least one processor, and the processor 11 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or the computer program loaded from the storage unit 18 to the random access memory (RAM) 13. In the RAM 13, various programs and data required for the operation of the electronic device can also be stored. The processor 11, the ROM 12, and the RAM 13 are connected to each other through a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0153] A number of components in the electronic device are connected to the I / O interface 15, including: an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a disk, an optical disk, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0154] The processor 11 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 11 executes the various methods and processes described above, such as a method for controlling the temperature of a water guide bearing of a hydro-turbine generator set.

[0155] In some embodiments, the water guide bearing temperature control method for a hydro-turbine generator set may be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as a storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed on an electronic device via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the water guide bearing temperature control method for a hydro-turbine generator set described above may be executed. Alternatively, in other embodiments, the processor 11 may be configured to execute the water guide bearing temperature control method for a hydro-turbine generator set in any other appropriate manner (for example, by means of firmware).

[0156] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0157] Computer programs for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that when the computer program is executed by the processor, the functions / operations specified in the flow chart and / or block diagram are implemented. The computer program may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0158] In the context of the present invention, a computer-readable storage medium may be a tangible medium that may contain or store a computer program for use by or in combination with an instruction execution system, device or equipment. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or equipment, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0159] To provide interaction with a user, the systems and techniques described herein may be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or trackball) through which the user can provide input to the electronic device. Other types of devices may also be used to provide interaction with the user; for example, the feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form (including acoustic input, voice input, or tactile input).

[0160] The systems and techniques described herein may be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0161] A computing system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The client and server relationship is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services.

[0162] An embodiment of the present invention further provides a computer program product, including a computer program, which, when executed by a processor, implements the method for controlling the temperature of a water guide bearing of a hydro-turbine generator set as provided in any embodiment of the present application.

[0163] In the process of implementation, the computer program product can be written in one or more programming languages ​​or a combination thereof to perform the computer program code of the present invention, including object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computer (for example, using an Internet service provider to connect through the Internet).

[0164] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps described in the present invention can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution of the present invention can be achieved, and this document does not limit this.

[0165] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for controlling the temperature of a water guide bearing of a hydro-turbine generator set, characterized in that: include: Collect historical temperature data of water guide bearings of hydro-turbine generator sets; Inputting the historical temperature data into a water-guide bearing temperature prediction model for processing, and predicting the predicted temperature of the water-guide bearing at the next control moment, wherein the water-guide bearing temperature prediction model integrates a convolutional neural network, a channel attention network, and a long short-term memory network; Based on the predicted temperature, model predictive control is used to optimize control parameters of the hydro-generator set.

2. The method for controlling the temperature of the water guide bearing of a hydro-turbine generator set according to claim 1, characterized in that: Before the historical temperature data is input into the water guide bearing temperature prediction model for processing, the method further includes: Performing data cleaning on the historical temperature data to remove abnormal data; Resample the cleaned historical temperature data to a fixed time interval; Perform linear interpolation on the resampled historical temperature data to align the data.

3. The method for controlling the temperature of the water guide bearing of a hydro-turbine generator set according to claim 1, characterized in that: Inputting the historical temperature data into the water-guide bearing temperature prediction model for processing, and predicting the predicted temperature of the water-guide bearing at the next control moment, includes: Inputting the historical temperature data into the convolutional neural network for processing to obtain local features of the historical temperature data; Inputting the local feature into the channel attention network for processing to obtain a channel attention weight vector, wherein the channel attention weight vector includes the channel attention weights of each channel of the local feature; Multiplying the local feature by the channel attention weight vector to obtain a channel attention feature; Inputting the channel attention feature into the long short-term memory network for processing to obtain the time series feature of the historical temperature data; The predicted temperature of the water guide bearing at the next control moment is predicted based on the time series characteristics.

4. The method for controlling the temperature of the water guide bearing of a hydro-turbine generator set according to claim 1, characterized in that: Based on the predicted temperature, the control parameters of the hydro-generator set are optimized by using model predictive control, including: Taking the unit load and the coolant level as inputs of the model predictive control, calculating the cost function value of the model predictive control; Determine whether the cost function value is greater than a threshold; If so, the unit load and the coolant level are updated based on the predicted temperature, the parameters of the model predictive control are adaptively learned, and the step of calculating the cost function value of the model predictive control by taking the unit load and the coolant level as the input of the model predictive control is returned to be executed until the cost function value is less than or equal to the threshold value, or the number of iterations reaches a preset number; If not, the control parameters of the hydro-generator set are output.

5. The method for controlling the temperature of the water guide bearing of a hydro-turbine generator set according to claim 4, characterized in that: Updating the unit load and the coolant level based on the predicted temperature includes: Calculate the Jacobian matrix of the water-guide bearing temperature prediction model based on the predicted temperature; Based on the Jacobian matrix, the unit load and the coolant level are updated using a gradient descent method.

6. The method for controlling the temperature of the water guide bearing of a hydro-turbine generator set according to claim 5, characterized in that: Based on the predicted temperature, the control parameters of the hydro-generator set are optimized by using model predictive control, which is converted into an optimization problem of the objective function: ∣Δu(t)∣≤Δu max ; u min ≤u(t)≤u max ; in, is the objective function, a and b are weight parameters, and the historical input is U p (t)=[u(t),u(t-1),…,u(tl u )], the historical output is Y p =[y(t),y(t-1),...,y(tl y )], and R(t)=[r(t+1),r(t+2),...,r(t+T p )] is the reference output, is the predicted output, U(t)=[u(t),u(t+1),...,u(t+T c -1)] is the optimal control input, ΔU(t)=[Δu(t),Δu(t+1),...,Δu(t+T c -1)] is the incremental control movement.

7. The method for controlling the temperature of the water guide bearing of a hydro-turbine generator set according to claim 6, characterized in that: Calculating the Jacobian matrix of the water guide bearing temperature prediction model based on the predicted temperature includes: The derivative of the output of the water-guided bearing temperature prediction model is calculated as the Jacobian matrix of the water-guided bearing temperature prediction model.

8. A water guide bearing temperature control device for a hydro-turbine generator set, characterized in that: include: A data acquisition module is used to collect historical temperature data of the water guide bearing of the hydro-turbine generator set; A temperature prediction module, used for inputting the historical temperature data into a water-guided bearing temperature prediction model for processing, and predicting the predicted temperature of the water-guided bearing at the next control moment, wherein the water-guided bearing temperature prediction model integrates a convolutional neural network, a channel attention network, and a long short-term memory network; A parameter optimization module is used to optimize the control parameters of the hydro-generator set by adopting model predictive control based on the predicted temperature.

9. An electronic device, characterized in that: include: one or more processors; A storage device for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the water guide bearing temperature control method of the hydro-turbine generator set as described in any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, a method for controlling the temperature of a water guide bearing of a hydro-turbine generator set as described in any one of claims 1 to 7 is implemented.

Images