A guide vane prediction method and device for optimizing efficiency of a hydraulic turbine

Abstract

This invention discloses a method and device for predicting guide vane efficiency optimization in hydropower generation, relating to the field of hydropower technology. The method includes: real-time collection of key parameters required for hydropower turbine operation using an industrial internet platform; preprocessing these parameters by outlier removal, filtering, and normalization to form a standardized dataset; calculating real-time efficiency based on the dataset; constructing a multi-condition efficiency characteristic database and model; and building a three-dimensional fitting surface for efficiency-head-flow rate. A machine learning algorithm is then employed, using key parameters as input and the optimal guide vane opening as output, embedding unit safety constraints, and performing cross-validation to train and optimize the prediction model. Finally, the current operating condition parameters are input, and the optimal opening recommendation value is output, maximizing turbine efficiency. This invention can adapt to changes in unit operating characteristics under different head conditions, significantly improving hydropower utilization.

Description

Technical Field

[0001] This invention relates to the field of hydropower technology, and in particular to a method and apparatus for predicting the efficiency of a water turbine guide vane. Background Technology

[0002] During the operation of a hydropower station, turbine efficiency directly affects power generation benefits and water resource utilization. Traditional turbine guide vane opening control mainly relies on empirical curves or fixed efficiency characteristic curves. The parameter settings of this type of control method are mostly based on the unit's factory test data or offline commissioning results, which are difficult to adapt to complex and changing inflow conditions, unit operating condition decline, and load fluctuations. Especially in run-of-river or mixed-flow power stations with large head variations, seasonal fluctuations in inflow can lead to frequent switching of unit operating conditions. Fixed opening strategies often cause the turbine to operate outside its optimal efficiency range for extended periods, resulting not only in water energy waste but also exacerbating unit vibration, cavitation, and other problems, thus shortening equipment lifespan.

[0003] In existing technologies, guide vane opening optimization methods are mostly based on theoretical efficiency characteristic curve derivation or numerical simulation analysis, which generally have obvious limitations: First, theoretical models and numerical simulations are difficult to accurately match the actual efficiency characteristics of actual units. Affected by factors such as differences in manufacturing processes, installation deviations, and operational wear, the actual efficiency curve deviates significantly from the theoretical curve. Second, the response timeliness is insufficient, and it is impossible to capture dynamic changes in operating conditions in real time and quickly adjust the opening parameters, resulting in prominent lag problems. Third, the consideration dimensions are singular, focusing only on basic parameters such as head and flow rate, failing to fully explore the efficiency correlation patterns contained in massive historical operating data, and the data value is not effectively utilized.

[0004] In summary, existing technologies are insufficient to meet the actual needs of safe, efficient, and economical operation of hydropower stations. There is an urgent need for an intelligent control method that can adapt to changes in operating conditions, deeply mine the value of operating data, and achieve precise optimization of guide vane opening. Summary of the Invention

[0005] The main objective of this invention is to provide a method for predicting the efficiency of a water turbine guide vane.

[0006] Another objective of this invention is to provide a guide vane prediction device for optimizing the efficiency of a water turbine.

[0007] The third objective of this invention is to provide an electronic device.

[0008] A fourth objective of this invention is to provide a non-transitory computer-readable storage medium.

[0009] To achieve the above objectives, a first aspect of the present invention proposes a method for predicting guide vane efficiency optimization in a hydraulic turbine, comprising:

[0010] S1, preprocess the collected key operating parameters of the water turbine to obtain a standardized dataset; S2 calculates the real-time operating efficiency of the turbine based on a standardized dataset, constructs a multi-condition structured efficiency characteristic database and efficiency characteristic model, and combines real-time efficiency values ​​with multi-dimensional operating parameters to complete the construction of a three-dimensional fitting surface of efficiency-head-flow rate. S3 employs a machine learning algorithm, taking the key operating parameters of the turbine in the standardized dataset as input features, and the guide vane opening corresponding to the optimal efficiency defined by the efficiency characteristic model and the three-dimensional fitting surface as output. The model parameters are optimized through cross-validation to complete the training and optimization of the optimal guide vane opening prediction model. S4 takes the active power, head, and unit flow rate under the current operating conditions, and after matching and verifying them with the efficiency characteristic database and the three-dimensional fitting surface, inputs them into the trained optimal guide vane opening prediction model. The model calculates and outputs the optimal guide vane opening suggestion value to maximize the operating efficiency of the turbine under the current operating conditions.

[0011] Optionally, the preprocessing of the collected key operating parameters of the water turbine to obtain a standardized dataset further includes: Through the industrial internet platform, the active power P, head H, unit flow rate Q, and guide vane opening Y of the water turbine are collected in real time during the operation of the turbine. Outlier removal is performed on the collected key operating parameters, deleting parameter data that is invalid or does not conform to the physical meaning of turbine operation; Noise filtering is performed on the key operational parameters after outlier removal. After noise filtering, the key operating parameters are normalized to form a standardized dataset.

[0012] Optionally, the step of calculating the real-time operating efficiency of the turbine based on a standardized dataset, constructing a structured efficiency characteristic database covering multiple operating conditions, simultaneously establishing an efficiency characteristic model, and combining real-time efficiency values ​​with multi-dimensional operating parameters to complete the construction of a three-dimensional fitting surface for efficiency-head-flow rate, further includes: Based on the energy conversion principle of a water turbine, the real-time operating efficiency of the water turbine is calculated using a thermodynamic efficiency formula. The specific calculation formula is as follows: ,in, To improve the real-time operating efficiency of the water turbine, Active power The density of water at room temperature and pressure. Where Q is the acceleration due to gravity, Q is the unit flow rate, and H is the head. The standardized dataset is categorized and stored according to operating condition type, including key operating parameters, corresponding real-time operating efficiency, and data collection timestamps. This creates a structured efficiency characteristic database, enabling orderly data management and rapid retrieval. Simultaneously establish an efficiency characteristic model and a three-dimensional fitting surface of efficiency-head-flow rate. The efficiency characteristic model is used to quantitatively characterize the dynamic law of efficiency changing with head, flow rate and guide vane opening. The three-dimensional fitting surface uses an intelligent algorithm to fit the correlation data of real-time operating efficiency with head and flow rate, and visualizes the coupling relationship of the three and the efficiency distribution range and efficiency peak region under different operating conditions.

[0013] Optionally, the process of training and optimizing the optimal guide vane opening prediction model includes: The XGBoost algorithm or LSTM algorithm is selected as the machine learning algorithm based on the characteristics of the actual working condition data. The active power P, head H, unit flow rate Q and guide vane opening Y in the standardized dataset are used as model inputs, and the guide vane opening that maximizes the real-time operating efficiency η of the turbine under the corresponding operating conditions is set as the model output. The unit vibration amplitude threshold, cavitation critical value, and guide vane opening mechanical limit value are used as model training constraints.

[0014] Optionally, the process of optimizing model parameters may also include: The standardized dataset is split into training and testing sets by using stratified sampling and according to a preset ratio. The model is trained using the training set and validated using the test set. The model hyperparameters were adjusted iteratively through multiple rounds of k-fold cross-validation, and the learning rate, number of iterations, and number of hidden layer nodes were optimized based on the mean squared error and coefficient of determination to complete the model parameter optimization.

[0015] Optionally, the proposed optimal guide vane opening value may further include: The active power, head and unit flow rate collected in real time under the current operating conditions are preprocessed using outlier removal, noise filtering and normalization methods to form standardized data of the current operating conditions that meet the model input requirements. The standardized data of the current operating condition is input into the trained optimal guide vane opening prediction model. Based on the learned efficiency rules and parameter coupling relationship, the model quickly completes inference calculation and outputs the optimal guide vane opening suggested value under the current operating condition. The optimal guide vane opening recommendation value is directly converted into a field-controllable operation command to guide the remote or manual adjustment of the turbine guide vanes, thereby maximizing the turbine's operating efficiency under the current operating conditions.

[0016] To achieve the above objectives, a second aspect of the present invention provides a guide vane prediction device for optimizing turbine efficiency, comprising: The data preprocessing module is used to preprocess the collected key operating parameters of the water turbine to obtain a standardized dataset; The efficiency characteristic modeling module is used to calculate the real-time operating efficiency of the turbine based on a standardized dataset, build a multi-condition structured efficiency characteristic database and efficiency characteristic model, and combine real-time efficiency values ​​with multi-dimensional operating parameters to complete the construction of a three-dimensional fitting surface of efficiency-head-flow rate. The model training and optimization module uses machine learning algorithms to take the key operating parameters of the turbine in the standardized dataset as input features, and the guide vane opening corresponding to the optimal efficiency defined by the efficiency characteristic model and the three-dimensional fitting surface as output. The module optimizes the model parameters through cross-validation to complete the training and optimization of the optimal guide vane opening prediction model. The optimal guide vane opening prediction module is used to input the active power, head, and unit flow rate under the current operating conditions into the trained optimal guide vane opening prediction model after matching and verifying the efficiency characteristic database with the three-dimensional fitting surface operating conditions. The model then generates and outputs the optimal guide vane opening suggestion value to maximize the operating efficiency of the turbine under the current operating conditions.

[0017] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0018] To achieve the above objectives, a third aspect of this application provides an electronic device, including a processor and a memory; wherein the processor reads executable program code stored in the memory to run a program corresponding to the executable program code, for implementing a guide vane prediction method for optimizing turbine efficiency as described in the first aspect embodiment.

[0019] To achieve the above objectives, the fourth aspect of this application provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a guide vane prediction method for optimizing turbine efficiency as described in the first aspect embodiment.

[0020] The embodiments of the present invention have the following beneficial effects: This invention adaptively matches the changing operating characteristics of the turbine under different head conditions, breaking the limitations of traditional fixed efficiency curves and experience-based control strategies. It accurately captures the efficiency optimization patterns under complex operating conditions, effectively guiding the turbine to always operate within its optimal efficiency range, significantly improving water resource utilization and reducing water waste. Furthermore, this method leverages the value of historical operating data mining, eliminating the need for theoretical models or numerical simulations. It can respond in real-time to dynamic changes in inflow conditions and turbine status, providing scientific, efficient, and intelligent decision support for the economic operation of hydropower stations. This reduces turbine operating losses, extends equipment lifespan, and achieves both economic and safety benefits. Attached Figure Description

[0021] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 A flowchart illustrating a method for predicting guide vane efficiency optimization for a water turbine, provided as an embodiment of the present invention; Figure 2 An overall framework diagram of a guide vane prediction method for optimizing turbine efficiency provided in an embodiment of the present invention; Figure 3 This is a structural diagram of the guide vane opening prediction model provided in an embodiment of the present invention; Figure 4 This is a partial set of unit operation data provided for embodiments of the present invention; Figure 5 This is a comparison chart of the optimal opening effect provided in the embodiments of the present invention; Figure 6 This is the optimized guide vane opening operation data for some units provided in the embodiments of the present invention; Figure 7 This is a schematic diagram of the adjusted optimized guide vane opening value and the measured guide vane value provided in an embodiment of the present invention; Figure 8 This is a structural diagram of a guide vane prediction device for optimizing the efficiency of a water turbine, provided in an embodiment of the present invention. Detailed Implementation

[0022] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0023] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0024] The following describes, with reference to the accompanying drawings, a method and apparatus for predicting guide vane efficiency optimization of a water turbine according to an embodiment of the present invention.

[0025] Example 1 This invention provides a method for predicting guide vane efficiency optimization in hydraulic turbines. Figure 1 This is a flowchart illustrating a method for predicting guide vane efficiency optimization in a water turbine, provided in an embodiment of the present invention. Figure 2 This is an overall framework diagram of a guide vane prediction method for optimizing turbine efficiency provided in an embodiment of the present invention. Figure 1 , Figure 2 As shown, the method includes the following steps: Step S1: Preprocess the collected key operating parameters of the water turbine to obtain a standardized dataset.

[0026] In this embodiment, the key parameters are precisely selected focusing on the core influencing dimensions of turbine efficiency assessment and turbine opening control, specifically covering four categories of indicators: active power P, head H, unit flow rate Q, and guide vane opening Y. To align with hydropower station on-site monitoring standards and practical engineering application needs, this application sets standardized units of measurement for each parameter. Active power P is measured in MW, head H in m, and unit flow rate Q in / h. This unified standard ensures the collected data is standardized and practical, allowing it to be directly used for subsequent calculations, analysis, and model training.

[0027] This application leverages the high-speed data transmission, distributed acquisition, and real-time monitoring capabilities of the industrial internet platform to achieve continuous and uninterrupted acquisition of the aforementioned key parameters. The acquisition frequency can be flexibly configured according to the complexity of the hydropower station's operating conditions. During the steady-state operation of the turbine, a conventional acquisition frequency can be used to meet basic monitoring needs; under dynamic conditions such as unit start-up and shutdown, and sudden load changes, it can automatically switch to a high-frequency acquisition mode to achieve accurate data capture of the changing operating conditions. Through this flexible acquisition strategy, it can comprehensively cover all operating scenarios, including turbine start-up, stable operation, load adjustment, and shutdown, effectively avoiding deviations in subsequent model training and efficiency analysis due to missing operating condition data, and ensuring the integrity and timeliness of the acquired data.

[0028] Because the raw data is susceptible to various factors during acquisition, such as sensor measurement accuracy, on-site electromagnetic interference, and equipment vibration, it inevitably contains invalid data, random noise, and differences in parameter magnitudes. Directly using this data for subsequent calculations and modeling will severely impact the accuracy of the results. Therefore, after data acquisition, a systematic preprocessing operation is necessary. This multi-stage, progressive processing eliminates various interference factors and refines data quality, providing reliable data support for subsequent efficiency calculations and model training. The specific preprocessing process is logically divided into three core steps: outlier removal, noise filtering, and normalization. In the outlier removal stage, invalid data with no physical meaning, such as active power P≤0 or unit flow rate Q≤0, are screened and removed. Simultaneously, based on the equipment's factory-rated operating parameter range, safe operating thresholds are defined, and extreme data exceeding these thresholds are removed, thus avoiding interference from invalid data on subsequent calculation accuracy from the source. In the noise filtering stage, a data smoothing algorithm is used to suppress noise. Specifically, by performing a moving average on continuously collected sequence data, random interference signals generated during sensor measurements are weakened, data fluctuations are reduced, and the processed data trend more closely matches the actual operating state of the equipment, significantly improving data stability. In the normalization stage, to address the issue of significant differences in the units and magnitudes of different parameters, a linear normalization method is used to uniformly map the values ​​of each parameter to the [0,1] interval, achieving parameter magnitude normalization and completely eliminating calculation biases caused by differences in units and magnitudes. This ultimately forms a high-precision, highly applicable standardized dataset, laying a solid data foundation for the smooth implementation of subsequent processes.

[0029] Step S2: Calculate the real-time operating efficiency of the turbine based on the standardized dataset, construct a structured efficiency characteristic database covering multiple operating conditions, simultaneously establish an efficiency characteristic model, and combine the real-time efficiency value with multi-dimensional operating parameters to complete the construction of the three-dimensional fitting surface of efficiency-head-flow rate.

[0030] In this embodiment, the calculation of the turbine's operating efficiency strictly follows the core principle of water energy to electrical energy conversion. A standardized formula is used to calculate the real-time efficiency value, ensuring the accuracy and authority of the efficiency calculation results at the engineering application level. The values ​​and definitions of each parameter fully conform to the actual application standards of hydropower engineering. The parameter *g* represents the density of water, fixed at 1000 kg / m³ under normal operating conditions at normal temperature and pressure; *g* represents gravitational acceleration, taken as 9.8 m / s² according to general engineering calculation standards; *P* is the active power collected in the S1 stage and preprocessed systematically; *Q* is the unit flow rate; and *H* is the head. All three parameters originate from standardized datasets and possess unified data quality standards. This formula can accurately quantify the turbine's water energy utilization efficiency under current operating conditions, intuitively and scientifically reflecting the effective degree of water energy to electrical energy conversion.

[0031] While performing real-time efficiency calculations, this application embodiment simultaneously constructs an efficiency characteristic database covering multiple operating conditions, providing a stable and reliable data carrier for historical operation data storage, full-condition process tracing, and subsequent predictive model training. This database adopts a highly scalable structured storage architecture, comprehensively storing both core and auxiliary data. It not only includes core indicator data such as active power, guide vane opening, and efficiency values ​​under different head and flow conditions, but also records detailed auxiliary information such as data acquisition time, operating condition type (e.g., startup, stable operation, load adjustment, shutdown), and equipment operating status (e.g., normal operation, slight vibration, moderate wear, maintenance pending), achieving full-dimensional, full-lifecycle data retention. Furthermore, the database supports precise data retrieval and rapid access based on multi-dimensional conditions such as parameter range, operating condition category, and acquisition time period, significantly improving data retrieval and usage efficiency, and achieving orderly management and efficient reuse of data across all operating conditions.

[0032] Simultaneously, this application embodiment establishes a turbine efficiency characteristic model. This model is an intelligent fitting model built based on multi-dimensional operating parameters. Its core function is to deeply explore the intrinsic correlation between parameters and efficiency, breaking the limitations of traditional empirical curves. In this application embodiment, the characteristic model specifically constructs a mapping relationship between multiple inputs and a single output. The input end consists of three core operating parameters: active power P, head H, and unit flow rate Q. The output end is the turbine efficiency. Its model structure adopts a classic three-layer neural network architecture, including an input layer, a hidden layer, and an output layer. The input layer is responsible for receiving standardized parameter data and completing data format conversion. The hidden layer deeply processes the nonlinear correlation between parameters through activation functions. The output layer outputs the efficiency prediction result. Through multiple rounds of iterative training on massive historical standardized data, the model weights and bias parameters are continuously optimized, ultimately obtaining the optimal model parameters. This enables the model to accurately capture the dynamic trend of efficiency changes with multi-dimensional parameters, providing solid quantitative support for efficiency law analysis.

[0033] Based on this, the embodiments of this application combine the efficiency value obtained from real-time calculation with multi-dimensional operating parameters such as active power, head, and flow rate, and generate a three-dimensional fitting surface of efficiency-head-flow rate using intelligent fitting algorithms (such as polynomial fitting, Gaussian process fitting, radial basis function fitting, etc.). This three-dimensional fitting surface can transform the abstract relationship between parameters and efficiency into an intuitive visual graph, clearly showing the efficiency distribution pattern under different head and flow rate combinations, accurately locating the efficiency peak region and the corresponding parameter range, providing clear and intuitive directional guidance for the search for the optimal guide vane opening in the subsequent S3 stage, helping to quickly lock the guide vane opening range corresponding to the optimal operating condition, and greatly improving the targeting and efficiency of subsequent model training.

[0034] Step S3: Using a machine learning algorithm, the key operating parameters of the turbine in the standardized dataset are used as input features. The optimal guide vane opening corresponding to the efficiency defined by the efficiency characteristic model and the three-dimensional fitting surface is used as the output. The model parameters are optimized through cross-validation to complete the training and optimization of the optimal guide vane opening prediction model.

[0035] In this embodiment, the machine learning algorithm selected is either the XGBoost algorithm or the LSTM algorithm. The two algorithms can be flexibly selected and adapted according to the characteristics of the actual operating data, taking into account the data adaptation requirements under different operating conditions. Among them, the XGBoost algorithm has a strong nonlinear fitting ability and is suitable for scenarios with strong nonlinear correlations between multidimensional operating parameters. It can deeply explore the complex correlations hidden between parameters and accurately capture the influence of parameter coupling on the guide vane opening. The LSTM algorithm is good at processing time-series data and is suitable for scenarios where the hydropower station operating conditions change dynamically over time and the data has significant time-series characteristics. It can effectively capture the time series patterns of operating condition changes and improve the model's adaptability to dynamic operating conditions.

[0036] During model training, such as Figure 3 As shown, the embodiments of this application strictly define the input features and output targets to ensure the relevance and effectiveness of model training. The input features are selected from the key operating parameters of the turbine after preprocessing in S1, specifically including active power P, head H, unit flow rate Q, and guide vane opening Y. After standardization, the above parameters can eliminate the influence of interference factors and provide high-quality data input for model training. The output target is clearly defined as the guide vane opening optimal_opening corresponding to the optimal efficiency. This optimal guide vane opening is determined by a formula, that is, within the preset feasible guide vane opening range, the guide vane opening value that can maximize the turbine efficiency is found by combining traversal search and optimization algorithms, ensuring that the model training target is highly consistent with the core requirement of maximizing turbine efficiency.

[0037] Meanwhile, to ensure the safe operation of the unit, this application's embodiments embed unit safety constraints during model training, constructing a dual-objective training mechanism of "optimal efficiency + controllable safety." These safety constraints specifically include core operational limitations such as unit vibration amplitude thresholds and cavitation thresholds. The vibration amplitude threshold is set based on equipment factory standards and on-site operating experience, while the cavitation threshold is dynamically adjusted in conjunction with operating parameters such as head and flow rate. By integrating these constraints into the model training process, the adjustment range of the guide vane opening is limited, preventing the model-predicted guide vane opening from exceeding the equipment's safe operating limits. This avoids equipment failures caused by improper opening at the source, balancing power generation efficiency and operational safety.

[0038] To further improve the model's prediction accuracy and stability, and effectively avoid overfitting or underfitting, this application employs cross-validation to optimize model parameters. Specifically, stratified sampling is used to divide the standardized dataset obtained in S1 into a training set and a test set at a preset ratio (e.g., 7:3). Stratified sampling ensures the consistency of the operating condition distribution between the training and test sets, avoiding the impact of data distribution bias on model performance evaluation. The training set is used for model parameter fitting training, iteratively calculating and optimizing core parameters such as model weights and biases. The test set is used to verify the model's generalization ability and prediction performance, and does not participate in the model parameter fitting process. Through multiple rounds of cross-validation iterations, the model's hyperparameters are adjusted. After each round of validation, the model's prediction error (e.g., mean squared error, mean absolute error), accuracy, and stability are quantitatively evaluated. Based on the evaluation results, the parameter configuration is dynamically optimized until all performance indicators of the model reach the preset standards. Finally, the training and parameter optimization of the optimal guide vane opening prediction model are completed, ensuring that the model can accurately adapt to the complex and ever-changing operating conditions of hydropower stations.

[0039] Step S4: The active power, head, and unit flow rate under the current operating conditions are matched and verified by the efficiency characteristic database and the three-dimensional fitting surface. Then, they are input into the trained optimal guide vane opening prediction model. The model is used to generate and output the optimal guide vane opening suggestion value, so as to maximize the operating efficiency of the turbine under the current operating conditions.

[0040] In this embodiment, preprocessing is first performed on the active power, head, and unit flow rate collected in real time under the current operating conditions. The preprocessing process is strictly consistent with S1, and three core steps are executed in sequence: outlier removal, noise filtering, and normalization. This ensures that the quality standards and standardization of the current input data are completely consistent with the data used for model training in the S3 stage, thereby eliminating prediction bias caused by differences in data processing procedures and providing high-quality and highly consistent input data support for the accurate calculation of the prediction model.

[0041] After preprocessing, the standardized parameters under the current operating conditions are input into the optimal guide vane opening prediction model, which has been trained and optimized in the S3 stage. Based on the parameter correlation patterns and efficiency optimization logic discovered in the early stage, this model can quickly complete data calculation and analysis, and finally generate the optimal guide vane opening recommendation value. This recommendation value has clear numerical accuracy and engineering applicability. It can be directly transmitted to the hydropower station's central control system, supporting the coordinated operation of automatic adjustment actuators. It can also be provided to on-site personnel to guide the manual adjustment of the guide vanes without the need for additional secondary conversion or correction, greatly improving the timeliness and efficiency of guide vane opening adjustment.

[0042] In this embodiment, based on the recommended value, operators can select automatic or manual control methods according to the actual control needs of the hydropower station, adjusting the guide vane opening in real time to enable the turbine to quickly adapt to the current inflow conditions and operating conditions, ensuring stable operation within the optimal efficiency range. This process not only maximizes the turbine's operating efficiency under current conditions and reduces ineffective water resource losses, significantly improving the power generation efficiency of the hydropower station, but also effectively reduces mechanical losses during unit operation, alleviates problems such as unit vibration and cavitation caused by unreasonable opening, reduces wear on equipment components, and extends the overall service life of the unit. Ultimately, this achieves multiple goals of economical, safe, and efficient operation of the hydropower station, providing strong technical support for its long-term stable operation.

[0043] In the application of one embodiment of the present invention, the implementation process is as follows: Application Example 1.

[0044] A large-scale mixed-flow hydropower station is used as an application example.

[0045] After the system of this application embodiment is started, the connection status of the industrial Internet platform of all hydropower units (hydropower unit 1, hydropower unit 2, ..., hydropower unit n) in the cascade basin is first initialized to ensure that the sensors, data transmission links and storage units of each unit are working normally.

[0046] (1) Based on the industrial internet platform for hydropower units in a cascade basin, the system collects and extracts the operating data of each unit in real time and in parallel, focusing on obtaining the following core parameters: The turbine generator head H is calculated from the water level difference measured by upstream and downstream water level gauges and is a key hydraulic parameter affecting the unit's energy conversion efficiency. The turbine guide vane opening Y is collected by the guide vane position sensor and directly determines the turbine's flow area and power output characteristics. Active power P of the unit: collected by a power transmitter, reflecting the current power generation output capacity of the unit; Unit flow rate Q: Calculated by flow sensor or volute differential pressure, representing the amount of water passing through the turbine per unit time.

[0047] The amount of data retrieved this time is approximately 12,000 records, such as... Figure 4 As shown, the collected raw data is first subjected to preliminary validity verification to remove erroneous data that is obviously outside the physical range, such as negative power values ​​or water head lower than the minimum operating water head of the unit, thus forming a structured raw dataset.

[0048] (2) Conduct systematic preprocessing of the raw data.

[0049] Step 1: Clean and standardize the raw data: The data cleaning process includes outlier removal, missing value handling, and deduplication to avoid invalid data interfering with subsequent calculations and modeling. Outlier removal uses the 3σ principle or box plot method to detect and remove abnormal data points caused by sensor drift, electromagnetic interference, or communication packet loss. Missing value handling uses linear interpolation or Lagrange interpolation to fill in a small amount of missing data in a short period of time. For missing data in a long period of time or a large amount of missing data, the time period is marked as invalid and removed. Deduplication removes duplicate data records caused by retransmission mechanism by comparing timestamps. The standardization process employs the Min-Max method, mapping parameters P, H, Q, and Y with different dimensions to the [0,1] interval. This eliminates dimensional differences, ensuring that different parameters have equal importance in subsequent model training and improving model stability and convergence speed. After processing, the standardized unit operation data {P,H,Q,Y} is output. The data acquisition, transmission, and preprocessing flow is as follows: Figure 2 As shown.

[0050] Step 2: Real-time computer group operating efficiency: The turbine efficiency η is calculated from the ratio of active power to hydraulic input power, and the mathematical formula is:

[0051] In the formula, ρ = 1000 kg / The density of water at normal temperature and pressure is given by g = 9.81 m³ / g. The standard value for gravitational acceleration is used. To ensure data validity, the calculated efficiency is filtered, retaining only samples under normal operating conditions that satisfy 0 < η ≤ 1. Based on this, the input features such as unit head H, flow rate Q, and guide vane opening Y are standardized using Z-score processing. The standardization formula is:

[0052] in, These are the original eigenvalues. This is the sample mean of this feature. The standard deviation of the samples is used to eliminate differences in dimensions and numerical magnitudes between different physical quantities through standardization, resulting in normalized features. This process generates a standardized feature dataset X and a turbine efficiency label set y, providing stable and standardized input data for subsequent turbine efficiency optimization models.

[0053] (3) Based on the standardized feature dataset X obtained after preprocessing and the turbine efficiency label set y, the efficiency prediction model is trained and evaluated. The first step is to divide the standardized dataset into a training set (8:2 ratio). , ) and test set ( , The training set is used for model fitting and training, and the test set is used for model performance verification. At the same time, the random seed is set to 42 to ensure that the dataset splitting results are reproducible, which facilitates subsequent experimental verification and method promotion, and avoids model performance fluctuations caused by random dataset splitting.

[0054] The second step involves constructing a random forest regression model as the efficiency prediction model, setting the number of decision trees to 100 and the random seed to 42, and using the training set data to complete the model's fitting training. Simultaneously, the system can also optimize a deep neural network (DNN) architecture, with maximizing unit efficiency η as the optimization objective: a mean squared error (MSE) loss function is constructed to measure the gap between predicted and actual efficiency, and the model weights are iteratively updated using the backpropagation algorithm; the dataset is divided into a training set (70%), a validation set (20%), and a test set (10%) to ensure the model's generalization ability; and hyperparameters such as the learning rate, the number of hidden layer nodes, and the activation function are adjusted through grid search or random search to prevent overfitting or underfitting, ultimately obtaining the most accurate guide vane opening prediction model.

[0055] After the model training is complete, the test set feature data will be used. Input the model to obtain the efficiency prediction value. The mean squared error (MSE) is used as the core evaluation index to quantify model performance, and its calculation formula is as follows: (where n is the number of samples in the test set, This represents the true efficiency value. (This refers to the predicted efficiency value). A smaller MSE value indicates that the model's predictions are closer to the true values, and the better the model's performance. The complete process of building, training, and evaluating the above efficiency prediction model is as follows: Figure 3 As shown.

[0056] (4) Under the given active power P of the unit, the unit head H (unit: meters), unit flow rate Q (unit: cubic meters per hour), guide vane opening Y (value range: 0~100%) and other operating data are collected by relying on the industrial Internet platform. The optimal guide vane opening value optimal_opening is predicted with the goal of maximizing the turbine efficiency, satisfying optimal_opening=f(P,H,Q,Y)=max(η).

[0057] The specific process is as follows: First, the training status of the efficiency prediction model is verified to determine whether the model has completed the complete training and evaluation process. If the model has not completed training or does not meet the usage conditions, an exception prompt is thrown directly and the subsequent optimization process is terminated to ensure the reliability and stability of the system operation. Then, an efficiency objective function η(H,Q,Y) with the guide vane opening Y as the independent variable is constructed. This function combines the head H, flow rate Q and guide vane opening Y under the current operating conditions to form a feature vector. After processing according to the standardization method described above, it is input into the trained efficiency prediction model. The turbine efficiency prediction value corresponding to this set of feature parameters is obtained through model inference.

[0058] Considering that conventional numerical optimization algorithms default to finding the minimum value as the optimization objective, to adapt to the algorithm's solution logic, the efficiency maximization problem is equivalently transformed into the minimization of inefficiency problem. η(H,Q,Y) is calculated, and a bounded scalar minimization method is used to find the optimal solution within the physical constraint of guide vane opening [0,100]%, ultimately obtaining the optimal guide vane opening. = And the maximum efficiency value at that opening degree. = (H,Q, ).

[0059] In practical applications, the system acquires the current operating data {P,H,Q} of the unit in real time and inputs it into the trained optimal guide vane opening prediction model. The model uses forward inference to find the opening value that maximizes the unit efficiency η among all possible guide vane openings, thus predicting the optimal guide vane opening (optimal_opening). The predicted optimal_opening is fed back to the main control system of the hydropower unit in real time as the target value for guide vane mechanism adjustment. The control system uses a PID controller or advanced control algorithm to drive a servo motor to precisely adjust the guide vane mechanism, enabling the actual guide vane opening to quickly track and stabilize near optimal_opening, ensuring that the unit operates at maximum efficiency under the current operating conditions.

[0060] In engineering applications, power plant test data containing active power P, head H, and flow rate Q (such as "Power Plant No. 6 Efficiency Test Data.csv") can be directly loaded to extract the head under specific operating conditions. ,flow Active power After substituting the values ​​into the optimization solution process, the optimal guide vane opening and corresponding maximum efficiency under this working condition can be output, and the results can be displayed intuitively in the form of "Optimal guide vane opening: XX.XX%".

[0061] The system establishes a continuous closed-loop mechanism: when the unit's operating conditions change (such as changes in head H or load demand P), it automatically repeats the entire process from data acquisition to opening adjustment; simultaneously, it periodically uses newly acquired operating data to incrementally train the prediction model, updating model parameters to adapt to operating condition drift caused by factors such as hydropower unit aging, changes in watershed hydrological conditions, and load demand fluctuations, ensuring the continuous effectiveness of prediction accuracy and achieving long-term optimization of the operating efficiency of hydropower units in the cascade watershed. The entire process ends when the system receives a shutdown command or meets preset termination conditions.

[0062] Through the above steps, a method for predicting the guide vane efficiency of a hydropower station turbine is completed.

[0063] Figure 5 The paper presents a comparison curve between the optimal guide vane opening value and the measured guide vane opening value, further illustrating the prediction of the optimal guide vane opening for maximizing the efficiency of the hydropower turbine. Figure 5 The system uses data such as active power P, turbine head H, and flow rate Q of the generating unit to predict the optimal guide vane opening for efficiency optimization. By acquiring relevant unit operating data and processing it through the guide vane prediction module, the optimal guide vane opening can be predicted. Figure 5 The red curve represents the measured value of the guide vane, and the blue curve represents the predicted value of the optimal guide vane opening. From Figure 5 In the embodiment, the active power P of the unit operates in the load range of 480MW-540MW, the measured opening of the guide vane varies in the range of 60%-85%, the predicted value of the optimal opening of the guide vane varies in the range of 45%-75%, and the measured value of the guide vane is about 10%-15% higher than the predicted value of the optimal opening of the guide vane.

[0064] Application Case 2.

[0065] According to a method for predicting the optimal guide vane opening of a turbine according to the present invention, the optimal guide vane opening result output by the optimal guide vane opening prediction model is used to adjust the guide vane opening value of the turbine in the actual project, so that the guide vane opening of the unit is close to the optimal guide vane prediction value. The following embodiment is the optimal guide vane opening data after adjustment.

[0066] Step 1: Based on the historical data storage unit of the Industrial Internet, acquire data such as the turbine generator head H, guide vane opening Y, active power P, and flow rate Q, obtaining approximately 14,000 data entries. Figure 6 As shown.

[0067] Step Two: Process the acquired unit operation data according to the above application examples (2), (3), and (4). Based on the efficiency model and guide vane optimization prediction model of this invention, calculate the optimal guide vane opening prediction value, such as... Figure 7 As shown.

[0068] Figure 7 The curve comparing the optimal guide vane opening value with the measured guide vane opening value in the specific implementation further illustrates the prediction of the optimal guide vane opening value for maximizing the efficiency of the hydropower turbine. Figure 7 The image shows the effect of processing data such as active power P, turbine head H, and turbine flow rate Q, with the turbine's operating parameters adjusted to the optimal guide vane opening. From... Figure 7 In this embodiment, the unit's active power P operates within a load range of 500MW-560MW. The measured guide vane opening varies between 64% and 83%, while the predicted optimal guide vane opening varies between 62% and 80%. The difference between the measured and predicted optimal guide vane opening values ​​is approximately 3%. By using the efficiency model and optimal guide vane prediction model based on this invention, the actual guide vane opening value can be effectively reduced, thereby reducing the unit's operating time. In engineering practice, adjusting the actual turbine guide vane opening to achieve the predicted optimal guide vane opening can meet the requirements under the current operating load, thereby reducing the unit's guide vane opening (reducing unit flow) and thus reducing unit losses.

[0069] Example 2 This invention provides a guide vane prediction device for optimizing the efficiency of a water turbine. Figure 8 This is a schematic flowchart of a guide vane prediction device for optimizing the efficiency of a water turbine, provided in an embodiment of the present invention. Figure 8 As shown, the device includes: The data preprocessing module 100 is used to preprocess the collected key operating parameters of the water turbine to obtain a standardized dataset; The efficiency characteristic modeling module 200 is used to calculate the real-time operating efficiency of the turbine based on a standardized dataset, construct a multi-condition structured efficiency characteristic database and efficiency characteristic model, and combine real-time efficiency values ​​with multi-dimensional operating parameters to complete the construction of a three-dimensional fitting surface of efficiency-head-flow rate. The model training and optimization module 300 is used to employ machine learning algorithms, taking the key operating parameters of the turbine in the standardized dataset as input features, and the guide vane opening corresponding to the optimal efficiency defined by the efficiency characteristic model and the three-dimensional fitting surface as output. The model parameters are optimized through cross-validation to complete the training and optimization of the optimal guide vane opening prediction model. The optimal guide vane opening prediction module 400 is used to input the active power, head and unit flow under the current operating conditions into the trained optimal guide vane opening prediction model after matching and verifying the efficiency characteristic database and the three-dimensional fitting surface operating conditions. The model calculates and outputs the optimal guide vane opening suggestion value to maximize the operating efficiency of the turbine under the current operating conditions.

[0070] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0071] Example 3 To implement the methods of the above embodiments, the present invention also provides an electronic device, which includes a memory and a processor; wherein the processor reads executable program code stored in the memory to run a program corresponding to the executable program code, so as to implement the various steps of the methods described above.

[0072] Example 4 To implement the above embodiments, this application also proposes a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the method described in the foregoing embodiments.

[0073] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0074] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0075] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

Claims

1. A method for predicting guide vane efficiency optimization in a hydraulic turbine, characterized in that, include: S1, preprocess the collected key operating parameters of the water turbine to obtain a standardized dataset; S2 calculates the real-time operating efficiency of the turbine based on a standardized dataset, constructs a multi-condition structured efficiency characteristic database and efficiency characteristic model, and combines real-time efficiency values ​​with multi-dimensional operating parameters to complete the construction of a three-dimensional fitting surface of efficiency-head-flow rate. S3 employs a machine learning algorithm, taking the key operating parameters of the turbine in the standardized dataset as input features, and the guide vane opening corresponding to the optimal efficiency defined by the efficiency characteristic model and the three-dimensional fitting surface as output. The model parameters are optimized through cross-validation to complete the training and optimization of the optimal guide vane opening prediction model. S4 takes the active power, head, and unit flow rate under the current operating conditions, and after matching and verifying them with the efficiency characteristic database and the three-dimensional fitting surface, inputs them into the trained optimal guide vane opening prediction model. The model calculates and outputs the optimal guide vane opening suggestion value to maximize the operating efficiency of the turbine under the current operating conditions.

2. The method according to claim 1, characterized in that, The process of preprocessing the collected key operating parameters of the water turbine to obtain a standardized dataset also includes: Through the industrial internet platform, the active power P, head H, unit flow rate Q, and guide vane opening Y of the water turbine are collected in real time during the operation of the turbine. Outlier removal is performed on the collected key operating parameters, deleting parameter data that is invalid or does not conform to the physical meaning of turbine operation; Noise filtering is performed on the key operational parameters after outlier removal. After noise filtering, the key operating parameters are normalized to form a standardized dataset.

3. The method according to claim 2, characterized in that, The process of calculating the real-time operating efficiency of a water turbine based on a standardized dataset, constructing a structured efficiency characteristic database covering multiple operating conditions, simultaneously establishing an efficiency characteristic model, and combining real-time efficiency values ​​with multi-dimensional operating parameters to complete the construction of a three-dimensional fitting surface for efficiency-head-flow rate, also includes: Based on the energy conversion principle of a water turbine, the real-time operating efficiency of the water turbine is calculated using a thermodynamic efficiency formula. The specific calculation formula is as follows: ,in, To improve the real-time operating efficiency of the water turbine, Active power The density of water at room temperature and pressure. Where Q is the acceleration due to gravity, Q is the unit flow rate, and H is the head. The standardized dataset is categorized and stored according to operating condition type, including key operating parameters, corresponding real-time operating efficiency, and data collection timestamps. This creates a structured efficiency characteristic database, enabling orderly data management and rapid retrieval. Simultaneously establish an efficiency characteristic model and a three-dimensional fitting surface of efficiency-head-flow rate. The efficiency characteristic model is used to quantitatively characterize the dynamic law of efficiency changing with head, flow rate and guide vane opening. The three-dimensional fitting surface uses an intelligent algorithm to fit the correlation data of real-time operating efficiency with head and flow rate, and visualizes the coupling relationship of the three and the efficiency distribution range and efficiency peak region under different operating conditions.

4. The method according to claim 3, characterized in that, The process of training and optimizing the optimal guide vane opening prediction model includes: The XGBoost algorithm or LSTM algorithm is selected as the machine learning algorithm based on the characteristics of the actual working condition data. The active power P, head H, unit flow rate Q and guide vane opening Y in the standardized dataset are used as model inputs, and the guide vane opening that maximizes the real-time operating efficiency η of the turbine under the corresponding operating conditions is set as the model output. The unit vibration amplitude threshold, cavitation critical value, and guide vane opening mechanical limit value are used as model training constraints.

5. The method according to claim 4, characterized in that, The process of optimizing model parameters also includes: The standardized dataset is split into training and testing sets by using stratified sampling and according to a preset ratio. The model is trained using the training set and validated using the test set. The model hyperparameters were adjusted iteratively through multiple rounds of k-fold cross-validation, and the learning rate, number of iterations, and number of hidden layer nodes were optimized based on the mean squared error and coefficient of determination to complete the model parameter optimization.

6. The method according to claim 5, characterized in that, The proposed optimal guide vane opening value also includes: The active power, head and unit flow rate collected in real time under the current operating conditions are preprocessed using outlier removal, noise filtering and normalization methods to form standardized data of the current operating conditions that meet the model input requirements. The standardized data of the current operating condition is input into the trained optimal guide vane opening prediction model. Based on the learned efficiency rules and parameter coupling relationship, the model quickly completes inference calculation and outputs the optimal guide vane opening suggested value under the current operating condition. The optimal guide vane opening recommendation value is directly converted into a field-controllable operation command to guide the remote or manual adjustment of the turbine guide vanes, thereby maximizing the turbine's operating efficiency under the current operating conditions.

7. A guide vane prediction device for optimizing the efficiency of a water turbine, characterized in that, include: The data preprocessing module is used to preprocess the collected key operating parameters of the water turbine to obtain a standardized dataset; The efficiency characteristic modeling module is used to calculate the real-time operating efficiency of the turbine based on a standardized dataset, build a multi-condition structured efficiency characteristic database and efficiency characteristic model, and combine real-time efficiency values ​​with multi-dimensional operating parameters to complete the construction of a three-dimensional fitting surface of efficiency-head-flow rate. The model training and optimization module uses machine learning algorithms to take the key operating parameters of the turbine in the standardized dataset as input features, and the guide vane opening corresponding to the optimal efficiency defined by the efficiency characteristic model and the three-dimensional fitting surface as output. The module optimizes the model parameters through cross-validation to complete the training and optimization of the optimal guide vane opening prediction model. The optimal guide vane opening prediction module is used to input the active power, head, and unit flow rate under the current operating conditions into the trained optimal guide vane opening prediction model after matching and verifying the efficiency characteristic database with the three-dimensional fitting surface operating conditions. The model then generates and outputs the optimal guide vane opening suggestion value to maximize the operating efficiency of the turbine under the current operating conditions.

8. The apparatus according to claim 7, characterized in that, The data preprocessing module is also used for: Through the industrial internet platform, the active power P, head H, unit flow rate Q, and guide vane opening Y of the water turbine are collected in real time during the operation of the turbine. Outlier removal is performed on the collected key operating parameters, deleting parameter data that is invalid or does not conform to the physical meaning of turbine operation; Noise filtering is performed on the key operational parameters after outlier removal. After noise filtering, the key operating parameters are normalized to form a standardized dataset.

9. An electronic device, characterized in that, Including processor and memory; The processor reads executable program code stored in the memory to run a program corresponding to the executable program code, so as to implement the method as described in any one of claims 1-6.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1-6.

Images