Hydraulic turbine governor damping analysis method and device, electronic equipment and storage medium

Abstract

The present disclosure provides a water turbine governor damping analysis method and device, electronic equipment and storage medium, relating to the technical field of hydroelectric power generation, by acquiring time series data of multiple monitoring parameters reflecting the operation state of the water turbine and the action of the governor, and inputting the time series data into a deep neural network with residual structure trained by a sample set constructed by historical operation data, and adjusted by a verified set, therefore, the problem of insufficient damping characteristic analysis precision and difficulty in adapting to complex operating conditions caused by relying on physical model simplification assumptions, lack of effective modeling of nonlinear dynamic correlation between multiple source monitoring parameters reflecting the operation state of the water turbine and the action of the governor, and lack of model precision guarantee through sample training and verification optimization based on historical data can be solved.

Description

Technical Field

[0001] This disclosure relates to the field of hydropower technology, and in particular to a damping analysis method and apparatus for a turbine governor, electronic equipment, and storage medium. Background Technology

[0002] The turbine speed control system is the core control unit of a hydropower station, widely used in power grid frequency regulation and ensuring the stability of unit operation. Within this technical framework, the industry has built a complete system covering dynamic modeling, feature extraction, and control optimization through the synergistic effect of hydraulic, mechanical, and electrical parameters. This system specifically includes key technical aspects such as physical modeling, mathematical simulation, and parameter identification. Among these, damping characteristic analysis is a core element ensuring the system maintains stable operation during frequency disturbance scenarios (such as primary and secondary frequency regulation processes).

[0003] As the demand for rapid frequency response capabilities in power grids continues to increase, traditional damping characteristic analysis methods are gradually showing their shortcomings in adaptability when dealing with multi-physics coupling problems involving hydraulic, mechanical, and electrical fields, as well as nonlinear dynamic modeling requirements.

[0004] Existing damping characteristic analysis methods generally rely on simplified assumptions based on physical models, such as linearizing relevant parameters, and lack in-depth correlation modeling of multi-source heterogeneous data such as roof vibration and water conduction sway. Furthermore, the dynamic coupling relationship between water flow disturbance and electromechanical parameters is difficult to capture effectively by traditional mathematical models, directly leading to a significant increase in the error of damping characteristic prediction results under operating conditions with sudden parameter changes. These technical limitations not only reduce the response speed of primary frequency regulation but may also cause the failure of secondary frequency regulation control strategies, resulting in frequency regulation delays exceeding two seconds. This significantly increases the safety risk of grid frequency collapse and further increases the operation and maintenance costs of the generating units. Summary of the Invention

[0005] This disclosure provides a method and apparatus for damping analysis of a turbine governor, as well as electronic equipment and a storage medium. Its main objective is to at least partially solve one of the technical problems in the related art.

[0006] According to a first aspect of this disclosure, a damping analysis method for a hydro turbine governor is provided, comprising: The timing data of multiple monitoring parameters of the water turbine during operation are acquired, and the monitoring parameters include at least those reflecting the operating status of the water turbine and the action of the governor. The time series data is input into a pre-trained damping characteristic analysis model, which is a deep neural network containing a residual structure, and is used to output analysis results characterizing the damping characteristics of the turbine speed regulation system based on the nonlinear dynamic correlation between the input multiple monitoring parameters. The damping characteristic analysis model is obtained through the following steps: A sample set is constructed based on historical operating data, and the sample set includes sample data of the multiple monitoring parameters and corresponding damping characteristic labels; The deep neural network containing the residual structure is trained using the sample set to minimize the difference between the predicted damping characteristics output by the network and the damping characteristic label, thereby obtaining a pre-trained damping characteristic analysis model. The trained model is evaluated using a validation set, and the model parameters or structure are adjusted based on the evaluation results until the model meets the preset accuracy requirements.

[0007] Optionally, acquiring time-series data of multiple monitoring parameters of the turbine during operation includes: Collect real-time data on active power, reactive power, guide vane opening, water guide sway, top cover horizontal vibration, and reservoir head; The collected real-time data is preprocessed, including removing outlier data and standardizing the data to eliminate the influence of units.

[0008] Optionally, the deep neural network containing the residual structure includes a plurality of sequentially connected residual units, each of the residual units including at least one convolutional layer and an addition operation connected to an identity mapping.

[0009] Optionally, training the deep neural network containing the residual structure using the sample set includes: The sample set is divided into a first sample subset for training and a second sample subset for validation. An optimization algorithm based on gradient descent is used to iteratively train the deep neural network with the first sample subset, and the training strategy is adjusted according to the performance on the second sample subset during the training process to prevent overfitting.

[0010] Optionally, adjusting the training strategy based on performance on the second sample subset includes: When the prediction error index of the damping characteristic analysis model on the second sample subset exceeds a preset threshold, the model complexity is adjusted by increasing the number of residual units of the deep neural network or introducing regularization constraints.

[0011] Optional, also includes: The damping characteristic analysis model that meets the preset accuracy requirements is deployed to the monitoring system of the water turbine; The deployed damping characteristic analysis model is periodically incrementally trained and its performance verified based on the latest operating data of the turbine, and the deployment damping characteristic analysis model is updated based on the verification results.

[0012] According to a second aspect of this disclosure, a damping analysis device for a water turbine governor is provided, comprising: The acquisition unit is used to acquire time-series data of multiple monitoring parameters of the turbine during operation, wherein the monitoring parameters include at least parameters reflecting the turbine's operating status and the governor's action; The output unit is used to input the time series data into a pre-trained damping characteristic analysis model, which is a deep neural network containing a residual structure, and is used to output analysis results characterizing the damping characteristics of the turbine speed regulation system based on the nonlinear dynamic correlation between the input multiple monitoring parameters. The damping characteristic analysis model is obtained through the following unit: A construction unit is used to construct a sample set based on historical operating data. The sample set includes sample data of the multiple monitoring parameters and corresponding damping characteristic labels. The training unit is used to train the deep neural network containing the residual structure using the sample set to minimize the difference between the predicted value of the damping characteristic output by the network and the damping characteristic label, so as to obtain a pre-trained damping characteristic analysis model. The evaluation unit is used to evaluate the trained model using a validation set and adjust the model parameters or structure based on the evaluation results until the model meets the preset accuracy requirements.

[0013] Optionally, the acquisition unit is also used for: Collect real-time data on active power, reactive power, guide vane opening, water guide sway, top cover horizontal vibration, and reservoir head; The collected real-time data is preprocessed, including removing outlier data and standardizing the data to eliminate the influence of units.

[0014] Optionally, the deep neural network containing the residual structure includes a plurality of sequentially connected residual units, each of the residual units including at least one convolutional layer and an addition operation connected to an identity mapping.

[0015] Optionally, the training unit is also used for: The sample set is divided into a first sample subset for training and a second sample subset for validation. An optimization algorithm based on gradient descent is used to iteratively train the deep neural network with the first sample subset, and the training strategy is adjusted according to the performance on the second sample subset during the training process to prevent overfitting.

[0016] Optionally, adjusting the training strategy based on performance on the second sample subset includes: When the prediction error index of the damping characteristic analysis model on the second sample subset exceeds a preset threshold, the model complexity is adjusted by increasing the number of residual units of the deep neural network or introducing regularization constraints.

[0017] Optional, also includes: The update unit is used to deploy the damping characteristic analysis model that meets the preset accuracy requirements to the monitoring system of the turbine; periodically perform incremental training and performance verification on the deployed damping characteristic analysis model based on the latest operating data of the turbine, and decide whether to update the deployed damping characteristic analysis model based on the verification results.

[0018] According to a third aspect of this disclosure, an electronic device is provided, comprising: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method described in the first aspect above.

[0019] According to a fourth aspect of this disclosure, a non-transitory computer-readable storage medium is provided storing computer instructions, wherein the computer instructions are configured to cause the computer to perform the method described in the first aspect above.

[0020] According to a fifth aspect of this disclosure, a computer program product is provided, comprising a computer program that, when executed by a processor, implements the method described in the first aspect above.

[0021] The damping analysis method, apparatus, electronic equipment, and storage medium for turbine governors disclosed herein acquire time-series data of multiple monitoring parameters reflecting the turbine's operating state and governor's actions. This time-series data is then input into a deep neural network with a residual structure, pre-trained using a sample set constructed from historical operating data and evaluated and adjusted using a validation set. Therefore, this method addresses the problems in existing technologies where reliance on simplified physical models, lack of effective modeling of the nonlinear dynamic correlation between multiple monitoring parameters reflecting the turbine's operating state and governor's actions, and failure to ensure model accuracy through sample training and validation optimization based on historical data lead to insufficient accuracy in damping characteristic analysis and difficulty in adapting to complex operating conditions. This method achieves the technical effect of accurately capturing the nonlinear dynamic correlation between multiple monitoring parameters, significantly improving the accuracy of damping characteristic analysis of the turbine speed control system, and providing reliable technical support for the stable operation of the turbine speed control system.

[0022] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0023] The accompanying drawings are provided to better understand this solution and do not constitute a limitation of this disclosure. Wherein: Figure 1 A flowchart illustrating a damping analysis method for a water turbine governor provided in this embodiment of the present disclosure; Figure 2 This is a schematic diagram of the structure of a damping analysis device for a water turbine governor provided in an embodiment of the present disclosure; Figure 3 A schematic block diagram of an example electronic device provided for embodiments of this disclosure. Detailed Implementation

[0024] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0025] The following description, with reference to the accompanying drawings, outlines an embodiment of a turbine governor damping analysis method and apparatus, electronic equipment, and storage medium.

[0026] Figure 1 This is a flowchart illustrating a damping analysis method for a water turbine governor provided in an embodiment of this disclosure.

[0027] like Figure 1As shown, the method includes the following steps: Step 101: Construct a sample set based on historical operating data. The sample set includes sample data of the multiple monitoring parameters and corresponding damping characteristic labels.

[0028] In the embodiments of this disclosure, to provide data support for the subsequent training of the damping characteristic analysis model, a sample set is first constructed based on historical operating data accumulated during the past operation of the turbine. This sample set must simultaneously include sample data of multiple monitoring parameters and corresponding damping characteristic labels to establish the correlation between monitoring parameters and damping characteristics. The historical operating data can cover relevant data collected during turbine operation under different conditions that reflect the system state. The multiple monitoring parameters need to be selected based on parameter types that characterize the turbine's operating state and governor's action. The damping characteristic labels are the actual damping characteristic representation information corresponding to the sample data, used as a target reference for model training. As one implementation method, the sample data of the multiple monitoring parameters can specifically include data on active power, reactive power, guide vane opening, water guide vane sway, top cover horizontal vibration, and reservoir head. The damping characteristic labels are determined based on the actual damping characteristic evaluation results of the above parameters during historical operation.

[0029] This sample set construction method ensures that a data foundation with clear input-output correlation is provided for model training, enabling the model to effectively learn the intrinsic relationship between monitoring parameters and damping characteristics. This provides data assurance for subsequent model training to achieve the preset accuracy, thereby supporting the accuracy of damping characteristic analysis of the turbine speed regulation system.

[0030] Step 102: The deep neural network containing the residual structure is trained using the sample set to minimize the difference between the predicted damping characteristic value output by the network and the damping characteristic label, thereby obtaining a pre-trained damping characteristic analysis model.

[0031] In the embodiments of this disclosure, to enable a deep neural network containing a residual structure to analyze the damping characteristics of a turbine speed regulation system, the network needs to be trained using the aforementioned constructed sample set. The training process aims to minimize the difference between the predicted damping characteristics output by the network and the corresponding damping characteristic labels in the sample set. Through continuous iterative adjustment of the network's internal parameters, the trained model is finally obtained. The deep neural network containing a residual structure can better handle the complex nonlinear relationships between multiple monitoring parameters and damping characteristics by relying on its residual connection characteristics, avoiding gradient correlation problems that may occur during deep network training. The objective of minimizing the difference provides a clear direction for adjusting the network parameters, ensuring that the model output gradually approaches the true damping characteristics. As one implementation method, the backpropagation algorithm can be used to update the network parameters during training, and a suitable optimizer can be set to assist in minimizing the difference. Specifically, the deep neural network containing the residual structure can be a ResNet network.

[0032] This training process allows the network to fully learn the correlation between the monitored parameters and damping characteristics in the sample set. The introduction of the residual structure helps the network capture complex dynamic relationships, while the difference minimization objective ensures the accuracy of model predictions, providing a reliable training foundation for subsequent model evaluation and practical applications.

[0033] Step 103: Use the validation set to evaluate the trained model, and adjust the model parameters or structure according to the evaluation results until the model meets the preset accuracy requirements.

[0034] In the embodiments of this disclosure, to ensure that the trained deep neural network containing residual structures (i.e., the damping characteristic analysis model) has reliable practical application capabilities, it is necessary to evaluate the trained model using a pre-prepared validation set. The core of the evaluation is to determine the degree of matching between the damping characteristic results output by the model based on the validation set and the corresponding real damping characteristics in the validation set, thereby verifying whether the model's generalization performance meets the standards. If the evaluation results show that the model does not meet the preset accuracy requirements, targeted adjustments need to be made to the model's core parameters (such as adjustment parameters that affect the network learning process) or basic structure (such as the number of network layers and the specific configuration of residual connections). After adjustment, the evaluation can be carried out again until the model's analytical accuracy fully meets the preset standards. As one implementation method, the validation set can be obtained from the historical operating data of the turbine and has no overlap with the training sample set. During the evaluation, the accuracy can be measured by indicators such as mean square error and root mean square error. During the adjustment process, the learning rate of the network, the batch size, or the depth of the residual structure can be optimized.

[0035] This evaluation and adjustment process can effectively avoid overfitting problems that may occur during model training, ensuring that the model maintains stable analytical accuracy in non-training data scenarios, and providing a reliable model guarantee for the accurate analysis of the damping characteristics of the turbine speed regulation system.

[0036] Step 104: Obtain time-series data of multiple monitoring parameters of the turbine during operation. The monitoring parameters include at least parameters reflecting the turbine's operating status and the governor's operation.

[0037] In the embodiments of this disclosure, to provide input data reflecting the dynamic operating state of the turbine to the subsequent damping characteristic analysis model, time-series data of multiple monitoring parameters need to be collected during the turbine's operation. This time-series data can completely record the changing patterns of each monitoring parameter over time, thereby reflecting the real-time evolution of the turbine's operating state and the governor's action response process. The selected multiple monitoring parameters must at least cover parameter types that characterize the turbine's operating state (such as power output, equipment operating stability, etc.) and the governor's actions (such as the operating state of regulating components, command execution, etc.) to ensure that the collected data can comprehensively correlate the core operating information of the turbine and the key action information of the governor, providing basic data support for the model's subsequent accurate analysis of damping characteristics. As one implementation method, the multiple monitoring parameters may specifically include active power, reactive power, guide vane opening, water guide vane sway, top cover horizontal vibration, and reservoir head. The collected time-series data correspondingly records the time-series changes of the above parameters during the turbine's operation.

[0038] This data acquisition method, by collecting time-series data and selectively choosing core monitoring parameters, ensures that dynamic and comprehensive input data is provided for the damping characteristic analysis model. This avoids analytical biases caused by static data or incomplete parameters, laying a reliable data foundation for the subsequent accurate output of damping characteristic analysis results by the model.

[0039] Step 105: Input the time series data into a pre-trained damping characteristic analysis model. The damping characteristic analysis model is a deep neural network containing a residual structure, which is used to output analysis results characterizing the damping characteristics of the turbine speed regulation system based on the nonlinear dynamic correlation between the input multiple monitoring parameters.

[0040] In the embodiments of this disclosure, to achieve effective analysis of the damping characteristics of the turbine speed regulation system, time-series data of multiple previously acquired monitoring parameters need to be input into a pre-trained damping characteristic analysis model. This damping characteristic analysis model is a deep neural network containing a residual structure. Its core function is to accurately identify and mine the nonlinear dynamic correlations between the multiple input monitoring parameters by relying on the residual structure's ability to process complex data relationships. Based on this correlation, it outputs analysis results that characterize the damping characteristics of the turbine speed regulation system, ensuring that the analysis results reflect the true state of the system's damping characteristics. As one implementation method, the deep neural network containing the residual structure can specifically be a ResNet network. The input time-series data can correspond to the time-series data of active power, reactive power, guide vane opening, water guide sway, top cover horizontal vibration, and reservoir head. The analysis results output by the model can be directly used to evaluate the damping state of the turbine speed regulation system.

[0041] This method uses a deep neural network of residual structure to mine the nonlinear dynamic correlation of monitoring parameters, which can effectively overcome the dependence of traditional methods on simplification assumptions, significantly improve the accuracy and reliability of damping characteristic analysis, and provide accurate technical basis for timely understanding of the stability of the turbine speed regulation system.

[0042] The damping analysis method for turbine governors disclosed herein acquires time-series data of multiple monitoring parameters reflecting the turbine's operating state and governor's actions. This time-series data is then input into a deep neural network with a residual structure, pre-trained using a sample set constructed from historical operating data and evaluated and adjusted using a validation set. Therefore, it addresses the problems in existing technologies where reliance on simplified assumptions based on physical models, lack of effective modeling of the nonlinear dynamic correlation between multiple monitoring parameters reflecting the turbine's operating state and governor's actions, and failure to ensure model accuracy through sample training and validation optimization based on historical data lead to insufficient accuracy in damping characteristic analysis and difficulty in adapting to complex operating conditions. This method achieves the technical effect of accurately capturing the nonlinear dynamic correlation between multiple monitoring parameters, significantly improving the accuracy of damping characteristic analysis of the turbine speed control system, and providing reliable technical support for the stable operation of the turbine speed control system.

[0043] As a specific implementation of this disclosure, based on the basic scheme, the acquisition of time-series data of multiple monitoring parameters of the turbine during operation is further defined, including: collecting real-time data of active power, reactive power, guide vane opening, water guide sway, top cover horizontal vibration and reservoir head; and preprocessing the collected real-time data, the preprocessing including removing abnormal data and standardizing the data to eliminate the influence of dimensions.

[0044] Specifically, based on the basic scheme of obtaining time-series data of multiple monitoring parameters during turbine operation, data is collected through dedicated sensors and monitoring devices mounted on the turbine operation monitoring system. Specifically, real-time active and reactive power data are calculated using current and voltage sensors in the power monitoring module; real-time guide vane opening data is collected using displacement sensors at the guide vane drive mechanism; real-time guide vane swing data is collected using eddy current sensors fixed to the outside of the turbine's guide vane bearings; real-time horizontal vibration data of the top cover is collected using piezoelectric accelerometers installed on the top cover; and real-time reservoir head data is collected using level sensors on the reservoir bank. The collection frequency is set to once per second to ensure the continuity of the time-series data. When preprocessing the above six types of real-time data, outliers are first removed using the 3σ principle, i.e., the mean μ and standard deviation σ of each type of data are calculated, and data exceeding the range of μ ± 3σ are identified as outliers and removed from the data sequence. Then, the Z-score standardization method is used to transform the data after outlier removal, according to the formula (x - The standardized data are obtained by calculating μ') / σ', where x is a single original data point, μ' is the mean of the data in this class after removing outliers, and σ' is the standard deviation of the data in this class after removing outliers. This eliminates the influence of different parameters due to differences in units (such as power in kW, opening in % and vibration in mm / s²).

[0045] Dedicated sensors are used to accurately collect core monitoring parameters, ensuring the authenticity and continuity of time-series data. The 3σ principle can effectively filter out abnormal data caused by sudden interference, and Z-score standardization can unify the data dimensions. The combination of the two provides high-quality input data for subsequent damping characteristic analysis models, avoiding data quality issues from affecting the accuracy of model analysis.

[0046] As a specific embodiment of this disclosure, based on the basic scheme, the deep neural network containing the residual structure is further defined to include a plurality of sequentially connected residual units, each of the residual units including at least one convolutional layer and an addition operation connected to an identity mapping.

[0047] Specifically, based on the deep neural network with residual structure described in the basic scheme, the network specifically includes 18 sequentially connected residual units (corresponding to the common ResNet-18 structure). Each residual unit is connected in series according to the feature extraction process, and the output feature map of the previous residual unit is directly used as the input of the next residual unit. Each residual unit contains two 3×3 convolutional layers and an identity mapping connection for addition. The first convolutional layer has 64 kernels (shallow layers) to 512 kernels (deep layers), with a stride of 1. The edge padding is set to "same" to ensure consistent input and output feature map sizes. After convolution, the feature maps undergo batch normalization and ReLU activation. The second convolutional layer has the same parameters as the first, except for the absence of an additional activation function. The identity mapping connection directly passes the original input feature map of the residual unit to the addition node. If the number of channels in the input feature map is different from the output feature map of the second convolutional layer, a 1×1 convolutional layer (with the same number of kernels and a stride of 1 as the second convolutional layer) is added in the identity mapping path to match the number of channels. Finally, the feature map of the identity mapping path is fused with the feature map output of the second convolutional layer through element-wise addition. The fused result is then activated by ReLU and used as the output of the residual unit.

[0048] By constructing a deep network through multiple sequentially connected residual units, it is possible to fully extract the feature information of the monitoring parameters through convolutional layers, solve the gradient vanishing problem in deep network training by using identity mapping, and retain the effective features of shallow layers through addition operations, thus providing reliable network structure support for accurately mining the nonlinear dynamic correlation between monitoring parameters.

[0049] As a specific embodiment of this disclosure, based on the basic scheme, the training of the deep neural network containing the residual structure using the sample set is further defined as follows: dividing the sample set into a first sample subset for training and a second sample subset for verification; using a gradient descent-based optimization algorithm to iteratively train the deep neural network with the first sample subset, and adjusting the training strategy according to the performance on the second sample subset during the training process to prevent overfitting.

[0050] Specifically, based on the basic scheme of training a deep neural network with residual structures using a sample set, the sample set is first divided into a first sample subset (training set) and a second sample subset (validation set) by randomly selecting samples in a 9:1 ratio. During the division, it is ensured that there are no overlapping samples between the two subsets, and that the working condition distribution and parameter range of the samples are consistent with the original sample set to avoid affecting the training and validation results due to data distribution deviations. The gradient descent-based optimization algorithm used for training is the Adam optimization algorithm, with an initial learning rate set to 0.001, a weight decay coefficient set to 1e-4 to constrain the parameter size, and a batch size configured to 32. The first sample subset... The deep neural network is trained iteratively. In each training round, the difference between the predicted value of the damping characteristic of the network output and the sample label is calculated using the mean squared error loss function. Then, the parameters of the network's convolutional layer, batch normalization layer, and other modules are updated using the backpropagation algorithm. During the training process, the network performance is evaluated using a second sample subset after every 4 rounds of iteration (the mean absolute error of the validation set is calculated). If the mean absolute error on the second sample subset does not decrease after 6 consecutive rounds of iteration (error fluctuation is less than 1e-5), the early stopping strategy is triggered to stop training. At the same time, L2 regularization is implemented through weight decay to adjust the training strategy and avoid the network overfitting the local features of the first sample subset.

[0051] By dividing the sample set in a 9:1 ratio, the sufficiency of training data and the representativeness of validation data are balanced. The Adam optimization algorithm can stably and efficiently update network parameters. Combined with early stopping strategy and L2 regularization, it can accurately control the training process based on the performance of the validation set, effectively suppress overfitting, significantly improve the model's generalization ability, and ensure that the model can still stably output accurate damping characteristic analysis results in practical applications.

[0052] As a specific implementation of this disclosure, based on the basic scheme, the method of adjusting the training strategy according to the performance on the second sample subset is further defined as follows: when the prediction error index of the damping characteristic analysis model on the second sample subset exceeds a preset threshold, the model complexity is adjusted by increasing the number of residual units of the deep neural network or introducing regularization constraints.

[0053] Specifically, based on the adjustment of the training strategy according to the performance of the second sample subset (i.e., the validation set) in the basic scheme, the prediction error index used for judgment is first clarified as the mean absolute percentage error (MAPE), and the preset threshold is set to 5%. That is, when the predicted value of the damping characteristic output by the model on the second sample subset exceeds 5% of the MAPE of the corresponding label, the training strategy adjustment is initiated. If the error exceeds the threshold due to insufficient model complexity, the original 18 residual units (corresponding to the ResNet-18 structure) of the deep neural network can be increased to 34 (corresponding to the ResNet-34 structure). The kernel size, number, and batch normalization configuration of the newly added residual units should be consistent with the original units, and they should be inserted in series into the original residual unit sequence to ensure that the feature extraction depth is improved while maintaining network structure compatibility. If the error exceeds the threshold due to model overfitting, regularization constraints should be introduced, specifically L1 regularization, adding an absolute value penalty term with a weight coefficient of 0.01 to the loss function, or adding a Dropout layer after the second convolutional layer of each residual unit of the network, setting the dropout probability to 0.2, and randomly blocking the output of some neurons to reduce model complexity and avoid over-reliance on noisy features in the training data.

[0054] By using clear error metrics and thresholds to achieve precise determination of adjustment triggers, adding residual units can specifically improve the depth of network feature extraction to adapt to complex data associations, while introducing regularization constraints can effectively suppress overfitting. The two adjustment methods can be selected as needed to quickly adjust the model complexity to a reasonable range to fit the data, ensuring that the model's prediction error on the second sample subset is reduced to within the threshold, thereby improving the model's generalization performance and reliability.

[0055] As a specific implementation of this disclosure, based on the basic scheme, the embodiment of this disclosure further includes: deploying the damping characteristic analysis model that meets the preset accuracy requirements to the monitoring system of the turbine; periodically performing incremental training and performance verification on the deployed damping characteristic analysis model based on the recent operating data of the turbine, and deciding whether to update the deployed damping characteristic analysis model based on the verification results.

[0056] Specifically, based on the basic scheme, a damping characteristic analysis model (i.e., a deep neural network containing residual structures) that meets the preset accuracy requirements is deployed to the turbine's monitoring system via an industrial-grade data interface (such as the OPC UA protocol interface). This allows the monitoring system to input time-series data of multiple monitoring parameters into the model in real time, and the damping characteristic analysis results output by the model are synchronously fed back to the monitoring system's visualization interface and early warning module, realizing real-time presentation of analysis results and anomaly triggering. A monthly cycle is set, and an incremental sample set is constructed based on the newly added operating data of the turbine during that cycle (covering six types of monitoring parameters, including active power and reactive power under different operating conditions, and corresponding damping characteristic labels). This incremental sample set is then merged with the sample data from the original sample set for the past six months. As a new training set, an incremental training method is adopted, which "freezes the parameters of shallow residual units and only fine-tunes the parameters of deep residual units and output layer" to avoid the waste of resources and increased training time caused by full training. After training, an independent recent validation set (20% randomly selected from the newly added data in the current month) is used to verify the model performance, with the mean absolute percentage error (MAPE) as the core indicator. If this indicator increases by more than 3% (preset threshold) compared with the model before deployment, the incrementally trained model will replace the model already deployed in the monitoring system; otherwise, the original model will remain unchanged.

[0057] Seamless integration of the model with the monitoring system is achieved through an industrial interface, ensuring the real-time nature of damping characteristic analysis. Periodic incremental training combined with fine-tuning of some parameters not only allows the model to adapt to new operating conditions and parameter changes of the turbine, but also reduces training costs. Based on clear indicators to determine whether to update the model, performance degradation caused by changes in data distribution can be effectively avoided, ensuring that the model has high-precision analysis capabilities in the long term, and providing continuous and reliable technical support for the stable operation of the turbine speed control system.

[0058] It should be noted that the embodiments of this disclosure may include multiple steps. For ease of description, these steps are numbered, but these numbers are not a limitation on the execution time slots or execution order between the steps; these steps can be implemented in any order, and the embodiments of this disclosure do not limit this.

[0059] Corresponding to the aforementioned damping analysis method for turbine governors, this disclosure also proposes a damping analysis device for turbine governors. Since the device embodiments of this disclosure correspond to the aforementioned method embodiments, details not disclosed in the device embodiments can be referred to the aforementioned method embodiments, and will not be repeated here.

[0060] Figure 2 This is a schematic diagram of the structure of a damping analysis device for a turbine governor provided in an embodiment of this disclosure, as shown below. Figure 2 As shown, it includes: The acquisition unit 21 is used to acquire time-series data of multiple monitoring parameters of the turbine during operation, wherein the monitoring parameters include at least parameters reflecting the turbine's operating status and the governor's action; Output unit 22 is used to input the time series data into a pre-trained damping characteristic analysis model, wherein the damping characteristic analysis model is a deep neural network containing a residual structure, and is used to output analysis results characterizing the damping characteristics of the turbine speed regulation system based on the nonlinear dynamic correlation between the input multiple monitoring parameters. The damping characteristic analysis model is obtained through the following unit: Construction unit 23 is used to construct a sample set based on historical operating data. The sample set includes sample data of the multiple monitoring parameters and corresponding damping characteristic labels. Training unit 24 is used to train the deep neural network containing the residual structure using the sample set, so as to minimize the difference between the predicted value of damping characteristics output by the network and the damping characteristic label, and obtain a pre-trained damping characteristic analysis model. Evaluation unit 25 is used to evaluate the trained model using a validation set and adjust the model parameters or structure according to the evaluation results until the model meets the preset accuracy requirements.

[0061] The turbine governor damping analysis device disclosed herein acquires time-series data of multiple monitoring parameters reflecting the turbine's operating state and governor's actions, and inputs this time-series data into a deep neural network with a residual structure, which has been trained using a sample set constructed from historical operating data and evaluated and adjusted using a validation set. Therefore, it solves the problems in existing technologies where reliance on simplified assumptions based on physical models, lack of effective modeling of the nonlinear dynamic correlation between multiple monitoring parameters reflecting the turbine's operating state and governor's actions, and failure to ensure model accuracy through sample training and validation optimization based on historical data lead to insufficient accuracy in damping characteristic analysis and difficulty in adapting to complex operating conditions. This device achieves the technical effect of accurately capturing the nonlinear dynamic correlation between multiple monitoring parameters, significantly improving the accuracy of damping characteristic analysis of the turbine speed control system, and providing reliable technical support for the stable operation of the turbine speed control system.

[0062] Furthermore, in one possible implementation of this embodiment, the acquisition unit 21 is also used for: Collect real-time data on active power, reactive power, guide vane opening, water guide sway, top cover horizontal vibration, and reservoir head; The collected real-time data is preprocessed, including removing outlier data and standardizing the data to eliminate the influence of units.

[0063] Furthermore, in one possible implementation of this embodiment, the deep neural network containing the residual structure includes a plurality of sequentially connected residual units, each of the residual units including at least one convolutional layer and an addition operation connected to an identity mapping.

[0064] Furthermore, in one possible implementation of this embodiment, the training unit 24 is also used for: The sample set is divided into a first sample subset for training and a second sample subset for validation. An optimization algorithm based on gradient descent is used to iteratively train the deep neural network with the first sample subset, and the training strategy is adjusted according to the performance on the second sample subset during the training process to prevent overfitting.

[0065] Furthermore, in one possible implementation of this embodiment, adjusting the training strategy based on the performance on the second sample subset includes: When the prediction error index of the damping characteristic analysis model on the second sample subset exceeds a preset threshold, the model complexity is adjusted by increasing the number of residual units of the deep neural network or introducing regularization constraints.

[0066] Furthermore, in one possible implementation of this embodiment, such as Figure 2 As shown, it also includes: The updating unit 26 is used to deploy the damping characteristic analysis model that meets the preset accuracy requirements to the monitoring system of the turbine; periodically perform incremental training and performance verification on the deployed damping characteristic analysis model based on the latest operating data of the turbine, and decide whether to update the deployed damping characteristic analysis model based on the verification results.

[0067] It should be noted that the foregoing explanation of the method embodiments also applies to the apparatus of this embodiment, and the principle is the same, so it is not limited in this embodiment.

[0068] According to embodiments of this disclosure, this disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0069] Figure 3A schematic block diagram of an example electronic device 300 that can be used to implement embodiments of the present disclosure is shown. 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 processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0070] like Figure 3 As shown, the electronic device 300 includes a computing unit 301, which can perform various appropriate actions and processes based on a computer program stored in ROM (Read-Only Memory) 302 or a computer program loaded from storage unit 308 into RAM (Random Access Memory) 303. The RAM 303 may also store various programs and data required for the operation of the electronic device 300. The computing unit 301, ROM 302, and RAM 303 are interconnected via a bus 304. An I / O (Input / Output) interface 305 is also connected to the bus 304.

[0071] Multiple components in electronic device 300 are connected to I / O interface 305, including: input unit 306, such as keyboard, mouse, etc.; output unit 307, such as various types of displays, speakers, etc.; storage unit 308, such as disk, optical disk, etc.; and communication unit 309, such as network card, modem, wireless transceiver, etc. Communication unit 309 allows electronic device 300 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0072] The computing unit 301 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 301 include, but are not limited to, CPUs (Central Processing Units), GPUs (Graphics Processing Units), various special-purpose AI (Artificial Intelligence) computing chips, various computing units running machine learning model algorithms, DSPs (Digital Signal Processors), and any suitable processors, controllers, microcontrollers, etc. The computing unit 301 performs the various methods and processes described above, such as the turbine governor damping analysis method. For example, in some embodiments, the turbine governor damping analysis method can be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as storage unit 308. In some embodiments, part or all of the computer program can be loaded and / or installed on the electronic device 300 via ROM 302 and / or communication unit 309. When the computer program is loaded into RAM 303 and executed by the computing unit 301, one or more steps of the methods described above can be performed. Alternatively, in other embodiments, the computing unit 301 may be configured to perform the aforementioned turbine governor damping analysis method by any other suitable means (e.g., by means of firmware).

[0073] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, FPGAs (Field Programmable Gate Arrays), ASICs (Application-Specific Integrated Circuits), ASSPs (Application-Specific Standard Products), SOCs (System-on-Chips), CPLDs (Complex Programmable Logic Devices), computer hardware, firmware, software, and / or combinations thereof. These various implementations may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0074] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0075] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, RAM, ROM, EPROM (Electrically Programmable Read-Only Memory) or flash memory, optical fiber, CD-ROM (Compact Disc Read-Only Memory), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0076] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (Cathode-Ray Tube) or LCD (Liquid Crystal Display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0077] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or computing systems that include middleware components (e.g., application servers), or computing systems that include frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include LANs (Local Area Networks), WANs (Wide Area Networks), the Internet, and blockchain networks.

[0078] Computer systems can include clients and servers. Clients and servers are generally geographically separated and typically interact via communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. A server can be a cloud server, also known as a cloud computing server or cloud host, a hosting product within the cloud computing service system that addresses the shortcomings of traditional physical hosts and VPS (Virtual Private Server) services, such as high management difficulty and weak business scalability. Servers can also be servers for distributed systems or servers incorporating blockchain technology.

[0079] It's important to note that artificial intelligence (AI) is the study of enabling computers to simulate certain human thought processes and intelligent behaviors (such as learning, reasoning, thinking, and planning). It encompasses both hardware and software technologies. AI hardware technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, and big data processing. AI software technologies primarily include computer vision, speech recognition, natural language processing, machine learning / deep learning, big data processing, and knowledge graph technologies.

[0080] The various numerical designations such as "first," "second," etc., used in this disclosure are merely for ease of description and are not intended to limit the scope of the embodiments of this disclosure, nor do they indicate a sequential order.

[0081] At least one of the features described in this disclosure can also be described as one or more, and multiple features can be two, three, four or more, and this disclosure does not impose any limitations. In the embodiments of this disclosure, for a technical feature, the technical features in that technical feature are distinguished by "first", "second", "third", "A", "B", "C" and "D", etc., and there is no sequential order or size order among the technical features described by "first", "second", "third", "A", "B", "C" and "D".

[0082] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0083] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A method for analyzing the damping of a turbine governor, characterized in that, include: The timing data of multiple monitoring parameters of the water turbine during operation are acquired, and the monitoring parameters include at least those reflecting the operating status of the water turbine and the action of the governor. The time series data is input into a pre-trained damping characteristic analysis model, which is a deep neural network containing a residual structure, and is used to output analysis results characterizing the damping characteristics of the turbine speed regulation system based on the nonlinear dynamic correlation between the input multiple monitoring parameters. The damping characteristic analysis model is obtained through the following steps: A sample set is constructed based on historical operating data, and the sample set includes sample data of the multiple monitoring parameters and corresponding damping characteristic labels; The deep neural network containing the residual structure is trained using the sample set to minimize the difference between the predicted damping characteristics output by the network and the damping characteristic label, thereby obtaining a pre-trained damping characteristic analysis model. The trained model is evaluated using a validation set, and the model parameters or structure are adjusted based on the evaluation results until the model meets the preset accuracy requirements.

2. The method according to claim 1, characterized in that, The acquisition of time-series data of multiple monitoring parameters of the water turbine during operation includes: Collect real-time data on active power, reactive power, guide vane opening, water guide sway, top cover horizontal vibration, and reservoir head; The collected real-time data is preprocessed, including removing outlier data and standardizing the data to eliminate the influence of units.

3. The method according to claim 1, characterized in that, The deep neural network containing residual structures comprises a plurality of sequentially connected residual units, each of which contains an addition operation connected to at least one convolutional layer and an identity mapping.

4. The method according to claim 1, characterized in that, The step of training the deep neural network containing the residual structure using the sample set includes: The sample set is divided into a first sample subset for training and a second sample subset for validation. An optimization algorithm based on gradient descent is used to iteratively train the deep neural network with the first sample subset, and the training strategy is adjusted according to the performance on the second sample subset during the training process to prevent overfitting.

5. The method according to claim 4, characterized in that, The step of adjusting the training strategy based on performance on the second sample subset includes: When the prediction error index of the damping characteristic analysis model on the second sample subset exceeds a preset threshold, the model complexity is adjusted by increasing the number of residual units of the deep neural network or introducing regularization constraints.

6. The method according to claim 1, characterized in that, Also includes: The damping characteristic analysis model that meets the preset accuracy requirements is deployed to the monitoring system of the water turbine; The deployed damping characteristic analysis model is periodically incrementally trained and its performance verified based on the latest operating data of the turbine, and the deployment damping characteristic analysis model is updated based on the verification results.

7. A damping analysis device for a water turbine governor, characterized in that, include: The acquisition unit is used to acquire time-series data of multiple monitoring parameters of the turbine during operation, wherein the monitoring parameters include at least parameters reflecting the turbine's operating status and the governor's action; The output unit is used to input the time series data into a pre-trained damping characteristic analysis model, which is a deep neural network containing a residual structure, and is used to output analysis results characterizing the damping characteristics of the turbine speed regulation system based on the nonlinear dynamic correlation between the input multiple monitoring parameters. The damping characteristic analysis model is obtained through the following unit: A construction unit is used to construct a sample set based on historical operating data. The sample set includes sample data of the multiple monitoring parameters and corresponding damping characteristic labels. The training unit is used to train the deep neural network containing the residual structure using the sample set to minimize the difference between the predicted value of the damping characteristic output by the network and the damping characteristic label, so as to obtain a pre-trained damping characteristic analysis model. The evaluation unit is used to evaluate the trained model using a validation set and adjust the model parameters or structure based on the evaluation results until the model meets the preset accuracy requirements.

8. An electronic device, characterized in that, include: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-6.

9. A non-transitory computer-readable storage medium storing computer instructions, characterized in that, The computer instructions are used to cause the computer to perform the method according to any one of claims 1-6.

10. A computer program product, characterized in that, Includes a computer program that, when executed by a processor, implements the method according to any one of claims 1-6.

Images