Method for determining operation efficiency of water feed pump turbine and monitoring system

Abstract

The invention provides a feed pump turbine operation efficiency determination method and a monitoring system, and relates to the technical field of turbine operation monitoring, and the method comprises the following steps: collecting operation data of a feed pump turbine; abnormal values in the operation data are removed; extracting data features in the cleaning data set, and performing dimension reduction on high-dimensional features in the data features; establishing an efficiency prediction model, and inputting historical operation data in the historical database into the model for training; calculating the actual operation efficiency of the feed pump turbine; and inputting the low-dimensional feature matrix into the trained efficiency prediction model, and predicting the operation efficiency of the water feed pump turbine. According to the method, the operation efficiency of the feed pump turbine is predicted by establishing the efficiency prediction model, shutdown is not needed, the production efficiency is not affected, possible operation efficiency reduction of the turbine during working can be found in time, and the reliability and safety of the system are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of steam turbine operation monitoring, and particularly to a method for determining the operation efficiency of a boiler feed pump steam turbine and a monitoring system. Background Art

[0002] Steam turbines, as a kind of thermal engine that converts steam thermal energy into mechanical energy, have a rather rich historical background and technical background. The development of steam turbines can be traced back to the 18th century. At that time, with the advancement of the Industrial Revolution, the demand for high-efficiency and high-power power was increasing day by day. The early steam turbine designs mainly focused on the UK and the US. The most famous one was the improved steam engine by James Watt. Although it was not a real steam turbine in the true sense, it laid the foundation for later steam turbine technologies. From the end of the 19th century to the beginning of the 20th century, with the progress of thermodynamics and materials science, real steam turbines began to appear. Charles Parsons in the UK and Carl Rust in Germany were two important figures in the history of steam turbine development. Parsons invented the first practical steam turbine in 1884, while Rust designed a more efficient steam turbine in 1904. These early steam turbines were mainly used to drive generators to provide power for industries and ships. Technically, the core principle of steam turbines is to convert the kinetic energy of steam into mechanical energy by using high-speed rotating blades. As the blade rotation speed increases, the efficiency and power density of steam turbines also increase significantly. The progress of materials science, especially the development of superalloys and heat-resistant coatings, enables steam turbines to operate at higher temperatures and pressures, thereby improving thermal efficiency and reliability. In addition, the application of computer-aided design and manufacturing (CAD / CAM) technologies makes the design and manufacturing of steam turbines more precise and efficient. Entering the 21st century, steam turbine technologies are still constantly progressing. With the increasing requirements for energy efficiency and environmental performance, steam turbine manufacturers are developing more efficient and cleaner power generation technologies.

[0003] The detection of steam turbine operation efficiency is an important link to ensure that steam turbines operate in the best state. Efficiency detection usually involves a comprehensive assessment of the thermal performance, mechanical performance, and electrical performance of steam turbines. Usually, by detecting a decrease in efficiency, fault diagnosis needs to be carried out to determine the root cause of the problem, such as blade fouling, wear of the flow path part, or control system failure. Necessary maintenance and adjustments are made according to the test results, such as cleaning the blades, replacing worn parts, or adjusting the control system, to improve the operation efficiency of the steam turbine.

[0004] For the detection of the operating efficiency of a general feed pump steam turbine, the performance test method is usually adopted, but this method has certain limitations and problems. The performance test may require the unit to be shut down, which affects production efficiency; the test results may be affected by various factors, such as operating condition changes, measurement errors, etc., resulting in inaccurate efficiency evaluation; in addition, the performance test usually only provides short-term efficiency data and is difficult to reflect the long-term operating trends and problems. Summary of the Invention

[0005] The present invention provides a method for determining the operating efficiency of a feed pump steam turbine and a monitoring system, so as to solve the limitations and problems existing in the prior art where the performance test method is usually adopted. The performance test may require the unit to be shut down, which affects production efficiency; the test results may be affected by various factors, such as operating condition changes, measurement errors, etc., resulting in inaccurate efficiency evaluation; in addition, the performance test usually only provides short-term efficiency data and is difficult to reflect the long-term operating trends and problems.

[0006] On the one hand, the present invention provides a method for determining the operating efficiency of a feed pump steam turbine, including: Collect the operating data of the feed pump steam turbine and align the time stamps of each item of operating data; Remove the outliers from the operating data and output the cleaned data set; Extract the data features from the cleaned data set and perform dimensionality reduction on the high-dimensional features in the data features, and output the low-dimensional feature matrix; Establish an efficiency prediction model and input the historical operating data in the historical database into the model for training; Calculate the actual operating efficiency of the feed pump steam turbine according to the operating data; Input the low-dimensional feature matrix into the trained efficiency prediction model to predict the operating efficiency of the feed pump steam turbine and output the predicted operating efficiency.

[0007] According to the method for determining the operating efficiency of a feed pump steam turbine provided by the present invention, the operating data includes steam parameters and pump parameters, and the steps for calculating the actual operating efficiency include: Calculate the steam inlet enthalpy value and the outlet enthalpy value according to the steam parameters; Calculate the input power of the steam turbine according to the inlet enthalpy value and the outlet enthalpy value; Calculate the effective output power of the feed pump according to the pump parameters; Calculate the actual operating efficiency according to the pump output power and the steam turbine input power.

[0008] According to the method for determining the operating efficiency of a feed pump steam turbine provided by the present invention, the steps for aligning the time stamps of each item of operating data include: Extract the time stamps corresponding to each parameter according to the steam parameters and the pump parameters; Refine the sampling frequency according to the timestamp; Compensate for the time offset according to the sampling frequency, including calculating the optimal offset using cross-correlation analysis; Divide the fixed time window according to the timestamp and take the latest value within the window value.

[0009] According to a method for determining the operating efficiency of a boiler feed pump steam turbine provided by the present invention, the steps of outputting a low-dimensional feature matrix include: Extract relevant features from the operating data according to the time window to generate a high-dimensional feature matrix; Extract an effective feature subset from the high-dimensional feature matrix; Perform dimensionality reduction processing on the effective feature subset to generate a low-dimensional feature matrix.

[0010] According to a method for determining the operating efficiency of a boiler feed pump steam turbine provided by the present invention, the steps of performing dimensionality reduction processing on the effective feature subset include: Eliminate the dimensional differences of the feature quantities in the feature subset, standardize the feature subset, and generate a data matrix; the standardization method includes calculating the mean and standard deviation of the feature subset and converting the data into a standard normal distribution with a mean of 0 and a standard deviation of 1; Create a projection matrix, and the matrix elements of the projection matrix are randomly filled from the standard normal distribution; Multiply the data matrix by the projection matrix to obtain a low-dimensional data matrix; Perform L2 normalization on the data in the low-dimensional data matrix; Calculate the mean square error between the data in the projected low-dimensional data matrix and the data in the feature subset to verify the degree of information loss.

[0011] According to a method for determining the operating efficiency of a boiler feed pump steam turbine provided by the present invention, the steps of establishing an efficiency prediction model include: Divide the historical operating data into a training set, a validation set, and a test set; Determine the objective function of the XGBoost model, and set the learning rate, the depth of the tree, and the maximum number of trees; Input the training set into the XGBoost model for training; when the depth of the tree reaches the preset depth, stop training; Evaluate the XGBoost model using the validation set; Optimize the depth of the trees of the XGBoost model using the goat optimization algorithm; Calculate the evaluation index of the XGBoost model using the test set.

[0012] According to a method for determining the operating efficiency of a boiler feed pump steam turbine provided by the present invention, the specific steps of optimizing the XGBoost model using the goat optimization algorithm include: Set the search space of the goat population; and initialize the goat population, where each goat individual in the population is a combination of hyperparameters; Calculate the fitness value of each goat individual; The goat individuals explore the search space by randomly moving and replace the initial positions with the new positions; The goat individuals gradually move towards the current optimal solution and continue to update their positions; The goats escape from the local optimal solution through a jumping mechanism; For the goat individuals with fitness values ranked lower than the preset value, randomly generate new positions for them; Continue to calculate the fitness value according to the new positions and update the current optimal solution; When the variance of the optimal solution in the goat population has no obvious change, stop updating the positions, and take the final optimal solution generated at the stop as the optimal hyperparameters of the XGBoost model.

[0013] According to a method for determining the operating efficiency of a feed water pump steam turbine provided by the present invention, it further includes: Compare the actual operating efficiency with the predicted operating efficiency, evaluate the efficiency prediction model, and output the evaluation result; Adaptively adjust the efficiency prediction model according to the evaluation result.

[0014] According to a method for determining the operating efficiency of a feed water pump steam turbine provided by the present invention, the ways to adaptively adjust the efficiency prediction model include: changing the number of features in the low-dimensional feature matrix to improve the performance of the efficiency prediction model; adjusting the parameters of the efficiency prediction model to improve the efficiency of the efficiency prediction model.

[0015] On the other hand, the present invention also provides a feed regulation system for beef cattle breeding, including: Sensors for real-time monitoring of various operating parameters of the feed water pump steam turbine, including steam parameters and water pump parameters; A data storage unit for storing the data transmitted by the sensors and converting it into digital signals; A central processing unit for processing and analyzing the data in the data storage unit and calculating the operating efficiency of the steam turbine; An intelligent control terminal for predicting the operating efficiency of the feed water pump steam turbine.

[0016] A method and monitoring system for determining the operating efficiency of a boiler feed pump steam turbine provided by the present invention can accurately calculate and predict the actual operating efficiency of the steam turbine by collecting and analyzing the operating data of the boiler feed pump steam turbine, including steam parameters and pump parameters. By establishing an efficiency prediction model to predict the operating efficiency of the boiler feed pump steam turbine, it is not necessary to stop the machine, thus not affecting the production efficiency, and it can timely detect the possible reduction in operating efficiency during the operation of the steam turbine, so as to timely investigate the reasons and improve the reliability and safety of the system. By using the goat optimization algorithm to optimize the efficiency prediction model, the accuracy and prediction efficiency of the prediction model are improved. By comparing the actual operating efficiency with the predicted operating efficiency to evaluate the efficiency prediction model, the deviation and deficiency of the efficiency prediction model can be timely detected and corrected, and the adaptability of the model can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0018] Figure 1 It is a flowchart of a method for determining the operating efficiency of a boiler feed pump steam turbine provided in Embodiment 1 of the present invention; Figure 2 It is a flowchart of predicting the operating efficiency for the efficiency prediction model; Figure 3 It is a schematic structural diagram of a monitoring system for the operating efficiency of a boiler feed pump steam turbine provided in Embodiment 2 of the present invention; DETAILED DESCRIPTION OF THE EMBODIMENTS

[0019] To make the objectives, technical solutions, and advantages of the present invention clearer, the following will clearly and completely describe the technical solutions in the present invention with reference to the drawings in the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art without creative efforts based on the embodiments in the present invention belong to the scope of protection of the present invention.

[0020] Embodiment 1 The following will describe Figure 1 - Figure 2 a method for determining the operating efficiency of a boiler feed pump steam turbine of the present invention.

[0021] As Figure 1 - Figure 2 shown, a method and monitoring system for determining the operating efficiency of a boiler feed pump steam turbine provided in an embodiment of the present invention specifically include the following steps: Collect the operation data of the feed water pump steam turbine. The operation data includes steam parameters and pump parameters. And align the timestamps of each operation data. This can ensure the accurate alignment of the operation data of the feed water pump and the steam turbine in the time dimension, providing a reliable basis for subsequent correlation analysis, fault prediction, and system optimization. The steps for aligning the timestamps of each operation data include: According to the steam parameters and pump parameters, the parameters include operation parameters such as the flow rate, pressure, and speed of the feed water pump, as well as parameters such as the steam temperature, pressure, flow rate, and speed of the steam turbine. Confirm the interfaces and storage formats (CSV, database, real-time stream) of the data acquisition system (such as SCADA, DCS, PI system). Extract the timestamps corresponding to each parameter. Check the timestamp format (such as Unix timestamp, ISO8601) and time zone settings. If the data comes from a cross-time zone system, it needs to be converted to a unified time zone (such as UTC+8). The timestamp accuracy usually needs to be accurate to milliseconds or microseconds.

[0022] Refine the sampling frequency according to the timestamps (such as once per second for the feed water pump and once per minute for the steam turbine). The specific methods include: interpolating the low-frequency data (such as steam temperature) to generate high-frequency timestamps. Performing linear interpolation on the parameters with steady changes (such as slow temperature changes). For step-type parameters (such as valve switch status), perform forward filling. When sampling high-frequency data, if the computational load needs to be reduced, the high-frequency data (such as the pressure of the feed water pump) can be aggregated. It is also possible to use the average value, maximum value, or sliding window statistics (such as calculating the average value every 5 seconds).

[0023] Compensate for the time offset according to the sampling frequency, including calculating the optimal offset using cross-correlation analysis. That is, due to different transmission delays from the sensor to the control system, the time offset needs to be calibrated. Usually, the optimal alignment offset needs to be calculated through data correlation (for example, the change in steam pressure causes a response delay in the feed water pump).

[0024] Divide fixed time windows according to the timestamps, and take the average value or the latest value within the window. Or use dynamic time warping to perform dynamic matching of the time series form to align the data.

[0025] Remove the outliers from the operation data and output the cleaned data set. For the operation of removing outliers, dynamic thresholds or isolation forest algorithms can be used to identify and remove the outlier points. Use neighboring data interpolation or LSTM-based time series prediction to fill in the missing values. Remove high-frequency noise through filters or wavelet transforms. The cleaned data is associated with the original data, and the data source and cleaning log are retained for the traceability of subsequent model training. The anomaly detection results are input into the subsequent feature engineering steps to generate anomaly feature labels, such as marking the abnormal time period.

[0026] Extract the data features from the cleaned dataset, perform dimensionality reduction on the high-dimensional features in the data features, and output the low-dimensional feature matrix. After feature engineering on the original data, a feature matrix containing statistical features, spectral features, and operating condition labels is generated. The reduced low-dimensional features are used as the input for the subsequent model to reduce the computational complexity. The steps for outputting the low-dimensional feature matrix include: According to the time window, extract relevant features from the operating data to generate a high-dimensional feature matrix. Specifically, it includes time feature extraction, frequency domain feature extraction, and operating condition feature extraction. When performing time feature extraction, calculate the sliding window statistics according to the time window, including mean difference, variance, maximum value, and minimum value, as well as skewness (distribution asymmetry) and kurtosis (distribution steepness). When performing frequency domain feature extraction, wavelet transform can be used for multi-scale decomposition to extract the energy of each layer of coefficients. For operating condition feature extraction, the operating conditions are usually divided based on the rotational speed or load threshold. For example, "high load" is defined as the rotational speed being greater than 3000 rpm (revolutions per minute). The features can also be combined and output. For example, "steam flow / current" reflects the unit energy consumption efficiency. The high-dimensional feature matrix generated after feature engineering extraction is usually a matrix containing a large number of sub-features.

[0027] Extract the effective feature subset from the high-dimensional feature matrix. Usually, the filtering method or the wrapper method is used to eliminate the features with low contribution or low correlation, and the effective feature subset is retained. The filtering method includes eliminating the constant features with variance close to 0 and calculating the Pearson correlation coefficient between the features and the efficiency label, and retaining the features with the absolute value of the coefficient greater than 0.3. The wrapper method refers to eliminating the features with low prediction contribution through the method of recursive feature elimination.

[0028] Perform dimensionality reduction on the effective feature subset to generate a low-dimensional feature matrix. The steps for performing dimensionality reduction include: eliminating the dimensionality differences of the feature quantities in the feature subset, standardizing the feature subset, and generating a data matrix. Data standardization is to eliminate the dimensionality differences between different features so that the contribution of each feature to the final projection result is uniform. This is crucial for maintaining the geometric structure of the data during the projection process. The standardization method includes calculating the mean and standard deviation of the feature subset and converting the data into a standard normal distribution with a mean of 0 and a standard deviation of 1. Calculate the mean (μ) and standard deviation (σ) for each feature column respectively. The formula is expressed as:

[0029]

[0030]

[0031] In the formula, is the standard normal distribution, x iis the feature subset, μ is the mean, σ is the standard deviation, i represents the i-th eigenvalue, and n is the number of features.

[0032] Create a projection matrix with the shape of d×k, where d is the original dimension and k is the target dimension. The matrix elements of the projection matrix are randomly filled from the standard normal distribution. Randomness ensures the uniform coverage of the projection and avoids systematic biases. And the value of k should satisfy k≥logn / ϵ 2 , where ϵ represents the allowable distance variance.

[0033] Multiply the data matrix (n×d) by the random projection matrix (d×k) to obtain the low-dimensional data matrix, denoted as: Z = X scaled ·R. In the formula, Z is the result of data compression in the data matrix, and its dimension is usually smaller than Xscaled. Xscaled is the standardized data, with a mean of 0 and a variance of 1 after eliminating the dimension difference. R is the transformation matrix, which is used to determine the direction and structure of the low-dimensional space. Through the linear (or non-linear) operation of the matrix R, the low-dimensional representation Z∈R n×d is obtained. By adjusting R, different task requirements can be adapted, such as the compression rate and interpretability. Usually, the eigenvalue decomposition of the covariance matrix needs to be calculated, and the first d eigenvectors are taken to learn R.

[0034] Perform L2 normalization on the data in the low-dimensional data matrix. It is expressed as:

[0035] In the formula, z j represents the feature after dimensionality reduction, and ||z j || 2 represents the L j norm of the vector z 2 , and z is the data after L2 normalization.

[0036] Calculate the mean square error between the data in the projected low-dimensional data matrix and the data in the feature subset to verify the degree of information loss.

[0037] Establish an efficiency prediction model and input the historical operation data in the historical database into the model for training.

[0038] Calculate the actual operation efficiency of the feedwater pump steam turbine according to the operation data.

[0039] The steam parameters include the inlet steam temperature and pressure, the outlet steam temperature and pressure, and the steam mass flow rate. The pump parameters include the pump output head, the pump flow rate, and the pump efficiency. The pump efficiency can be obtained from the equipment manual or calculated through actual tests. The steps for calculating the actual operation efficiency include: Calculate the inlet enthalpy value and the outlet enthalpy value of the steam according to the steam parameters.

[0040]

[0041] In the formula, P is the steam pressure, T is the steam temperature, and h is the enthalpy value of the system, that is, the enthalpy value at the inlet or outlet. h ref is the reference enthalpy value, usually a known reference point. T ref is the reference absolute steam temperature, c p is the specific heat capacity at constant pressure, indicating the heat required for the system temperature to rise by 1 K under constant pressure. represents the integral of the specific heat capacity at constant pressure from the reference temperature T ref to the current temperature T, that is, the heat required to heat the system from the reference temperature to the current temperature. v is the steam volume, P ref is the absolute steam pressure under the reference, and v(P−P ref ) represents the change in the enthalpy of the system due to the pressure change.

[0042] Calculate the input power of the steam turbine according to the inlet enthalpy value and the outlet enthalpy value. The formula is expressed as:

[0043] In the formula, Q in represents the input power of the steam turbine, m s is the steam mass flow rate, h 1 is the steam inlet enthalpy value, h 2 is the steam outlet enthalpy value.

[0044] Calculate the effective output power of the feed pump according to the pump parameters. Usually, consider the water density ρ≈1000 kg / m3 and the gravitational acceleration g = 9.81 m / s2. The formula is expressed as:

[0045] In the formula, P out represents the effective output power, and H represents the head of the feed pump.

[0046] Calculate the actual operating efficiency according to the pump output power and the steam turbine input power. The formula is expressed as:

[0047] In the formula, η represents the actual operating efficiency, and α represents the mechanical transmission efficiency, and the value is usually 95% - 98%.

[0048] Input the low-dimensional feature matrix into the trained efficiency prediction model to predict the operating efficiency of the feed pump steam turbine, and output the predicted operating efficiency.

[0049] The steps to establish the efficiency prediction model include: Divide the historical operation data into a training set, a validation set, and a test set. The splitting ratio follows the 7:1:2 principle.

[0050] Determine the objective function of the XGBoost model, and set the learning rate, the depth of the tree, and the maximum number of trees. The formula of the objective function is expressed as:

[0051]

[0052] In the formula, is the objective function, is the loss function, y a represents the true label value of the a-th sample, and a is the sample serial number. is the predicted value of the a-th sample. θ is the learning rate, and its value range is (0,1]. By reducing the contribution of a single tree update, the model is forced to learn more trees. U is the maximum tree depth, γ is the Gini coefficient, that is, the pruning parameter, and its value range is [0,1], which is used to control the penalty intensity of the model complexity. When γ increases, the model is more inclined to select simpler trees (reduce the splitting nodes). m is the total number of samples in the training set. w g represents the weight of the tree, that is, the feature weight vector of the g-th decision tree. α 1 is the L 2 regularization coefficient, α 1 ≥0, which is used to control the smoothness of the model. It is usually used to impose a penalty on the sum of squares of the tree weights, prompting the model to select a smoother decision boundary. Compared with L1 regularization, L2 is more inclined to evenly disperse the influence of outliers to all relevant features.

[0053] Input the training set into the XGBoost model for training. During the training process, the parameters of the model will be continuously adjusted to minimize the loss function and improve the prediction accuracy of the model. When the depth of the tree reaches the preset depth, stop training. This is to avoid overfitting, that is, the model performs well on the training set but poorly on unknown data. By restricting the depth of the tree, the generalization ability of the model can be improved, making it have better prediction performance when facing unknown data.

[0054] Use the validation set to evaluate the XGBoost model. By calculating the prediction error and relevant evaluation metrics on the validation set, the performance of the model on unknown data can be understood, so as to judge whether the model is overfitting or underfitting.

[0055] Optimize the depth of the trees in the XGBoost model using the Goat Optimization Algorithm. The Goat Optimization Algorithm is a swarm intelligence-based optimization algorithm that finds the optimal solution by simulating the foraging behavior of goats. During the optimization process, the algorithm automatically adjusts the depth of the trees based on the performance metrics of the model to find the best model parameters.

[0056] The specific steps to optimize the XGBoost model using the Goat Optimization Algorithm are as follows: Set the search space for the goat population. And initialize the goat population, where each goat individual in the population is a combination of hyperparameters. The formula is expressed as:

[0057]

[0058] In the formula, r is the number of decision variables, that is, the dimension of the search space. k is the index of each goat individual in the search space, and B k is the position of the k-th goat, expressed as an r-dimensional vector, LB is the lower bound of each dimension in the search space, UB is the upper bound of each dimension in the search space, and rand(r) represents generating an r-dimensional random vector, with the value of each component between [0,1].

[0059] Calculate the fitness value of each goat individual. The formula is expressed as:

[0060] In the formula, MSE is the fitness function, n 1 is the total number of goats in the population, y 1k is the true position of the goat individual, is the predicted position of the goat individual.

[0061] Goat individuals explore the search space by randomly moving and replace the initial position with the new position. The formula is expressed as:

[0062] In the formula, is the new position of the k-th goat, is the position of the k-th goat at iteration t, δ is the exploration coefficient, and R is a random variable drawn from the Gaussian distribution (0,1).

[0063] Goat individuals gradually move towards the current optimal solution and continue to update their positions to refine the quality of the solution. The formula is expressed as:

[0064] In the formula, is the current optimal solution, and δ' is the exploitation coefficient.

[0065] The goat escapes from the local optimum through a jumping mechanism. The position update formula is expressed as:

[0066] In the formula, J is the jumping coefficient, is the position of the random goat.

[0067] For goat individuals with fitness values ranked lower than the preset value, a new position is randomly generated for them, and the reset formula is:

[0068] By resetting its position to a newly randomly generated position, the diversity and robustness of the population are maintained.

[0069] Continue to calculate the fitness value according to the new position and update the current optimal solution.

[0070] When the variance of the optimal solution in the goat population does not change significantly, stop updating the position, and take the final optimal solution generated at the stop as the optimal hyperparameter of the XGBoost model.

[0071] Use the test set to calculate the evaluation metrics of the XGBoost model. By calculating the prediction error and related evaluation metrics on the test set, the performance of the model in practical applications can be obtained. These evaluation metrics include but are not limited to accuracy, recall rate, F1 score, etc.

[0072] The method for determining the operating efficiency of the feed water pump steam turbine further includes: comparing the actual operating efficiency with the predicted operating efficiency, evaluating the efficiency prediction model, and outputting the evaluation result.

[0073] According to the evaluation result, perform adaptive adjustment on the efficiency prediction model.

[0074] The ways to perform adaptive adjustment on the efficiency prediction model include: changing the number of features in the low-dimensional feature matrix to improve the performance of the efficiency prediction model. Adjusting the parameters of the efficiency prediction model to improve the efficiency of the efficiency prediction model.

[0075] Embodiment 2: As Figure 3 shown, on the other hand, the present invention also provides a feed regulation system for beef cattle breeding, including: Sensors for real-time monitoring of various operating parameters of the feed water pump steam turbine, including steam parameters and water pump parameters. These parameters may include but are not limited to the temperature, pressure, and flow rate of the steam, as well as the rotational speed, flow rate, and head of the water pump. The types of sensors may include thermocouples, pressure sensors, flow meters, etc., which can convert these physical quantities into electrical signals for further processing and analysis.

[0076] A data storage unit is used to store the data transmitted by sensors and convert it into digital signals so that the data can be recognized and processed by the central processing unit. The data storage unit usually includes an analog-to-digital converter (ADC) and sufficient storage space to save historical data for subsequent analysis and fault diagnosis.

[0077] The central processing unit is used to process and analyze the data in the data storage unit and calculate the operating efficiency of the steam turbine. It usually includes one or more microprocessors and necessary software algorithms for calculating the operating efficiency of the steam turbine. These algorithms may be based on physical models, statistical analysis, or machine learning techniques, capable of extracting useful information from sensor data and calculating the performance indicators of the steam turbine.

[0078] The intelligent control terminal is used to predict the operating efficiency of the feed water pump steam turbine. The intelligent control terminal contains an automatic optimization algorithm that can predict the operating efficiency of the feed water pump steam turbine through a model and can automatically adjust the operating parameters of the steam turbine according to the prediction results to improve its efficiency and reliability.

[0079] Example 1: The operating parameters of the system where the steam turbine drives the water pump are as follows: the inlet steam temperature is 450 °C, the inlet steam pressure is 4 MPa, the outlet steam temperature is 152 °C, the outlet steam pressure is 0.5 MPa, and the steam mass flow rate is 80 kg / s. The head of the feed water pump is 50 m, the flow rate is 30 m³ / s, and the efficiency of the water pump is found to be 85%. Thus, by querying the steam enthalpy value table or through calculation, the inlet enthalpy value of the steam can be obtained as 3300 kJ / kg, and the outlet enthalpy value is 2748 kJ / kg. From this, the input heat Q in of the steam turbine can be calculated to be 44160 kW. Substituting the head and flow rate of the water pump into the formula, the effective output power P out of the water transfer pump can be calculated to be 16495 kW. The mechanical transmission efficiency is set at 95%. Finally, based on the input heat Q in of the steam turbine and the effective output power P out of the water pump, the operating efficiency of the steam turbine is calculated to be 35.5%.

[0080] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative labor.

[0081] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on such an understanding, the essence of the above technical solution, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0082] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for determining the operating efficiency of a feedwater pump turbine, characterized in that: include: Collecting the operating data of the feedwater pump turbine and aligning the timestamps of each item of the operating data; removing outliers in the operating data and outputting a cleaned data set; Extracting data features from the cleaned data set, and reducing the dimensions of high-dimensional features in the data features to output a low-dimensional feature matrix; Establish an efficiency prediction model and input historical operation data from the historical database into the model for training; Calculating the actual operating efficiency of the feedwater pump turbine according to the operating data; The low-dimensional feature matrix is ​​input into the trained efficiency prediction model to predict the operating efficiency of the feedwater pump turbine and output the predicted operating efficiency.

2. A method for determining the operating efficiency of a feedwater pump turbine according to claim 1, characterized in that: The operation data includes steam parameters and water pump parameters, and the step of calculating the actual operation efficiency includes: Calculating steam inlet enthalpy and outlet enthalpy according to the steam parameters; Calculating the turbine input power according to the inlet enthalpy value and the outlet enthalpy value; Calculate the effective output power of the water pump according to the water pump parameters; The actual operation efficiency is calculated according to the water pump output power and the steam turbine input power.

3. A method for determining the operating efficiency of a feedwater pump turbine according to claim 2, characterized in that: The steps of aligning the timestamps of the various operation data include: According to the steam parameter and the water pump parameter, extracting the timestamp corresponding to each parameter; Refining the sampling frequency according to the timestamp; Compensating for time offsets based on sampling frequency, including calculating an optimal offset using cross-correlation analysis; A fixed time window is divided according to the timestamp, and the latest value in the window is taken.

4. A method for determining the operating efficiency of a feedwater pump turbine according to claim 3, characterized in that: The steps to output a low-dimensional feature matrix include: Extract relevant features from the operation data according to the time window to generate a high-dimensional feature matrix; Extracting a valid feature subset from the high-dimensional feature matrix; Perform dimensionality reduction processing on the effective feature subset to generate a low-dimensional feature matrix.

5. A method for determining the operating efficiency of a feedwater pump turbine according to claim 4, characterized in that: The step of performing dimensionality reduction processing on the effective feature subset comprises: Eliminating the feature dimension differences in the feature subset, standardizing the feature subset, and generating a data matrix; the standardization method includes calculating the mean and standard deviation of the feature subset, and converting the data into a standard normal distribution with a mean of 0 and a standard deviation of 1; Creating a projection matrix, wherein the matrix elements of the projection matrix are randomly filled from the standard normal distribution; Multiplying the data matrix by the projection matrix to obtain a low-dimensional data matrix; Performing L2 normalization on the data in the low-dimensional data matrix; The mean square error between the data in the low-dimensional data matrix after projection and the data in the feature subset is calculated to verify the degree of information loss.

6. A method for determining the operating efficiency of a feedwater pump turbine according to claim 1, characterized in that: The steps to build an efficiency prediction model include: Dividing the historical operation data into a training set, a validation set and a test set; Determine the objective function of the XGBoost model, set the learning rate, tree depth, and maximum number of trees; Input the training set into the XGBoost model for training; when the depth of the tree reaches a preset depth, stop training; Evaluate the XGBoost model using the validation set; Optimizing the depth of the tree of the XGBoost model using the goat optimization algorithm; The evaluation metrics of the XGBoost model are calculated using the test set.

7. A method for determining the operating efficiency of a feedwater pump turbine according to claim 6, characterized in that: The specific steps to optimize the XGBoost model using the goat optimization algorithm include: Setting a search space of a goat population; and initializing the goat population, wherein each goat individual in the population serves as a hyperparameter combination; Calculating the fitness value of each of the goat individuals; The goat individuals explore the search space by randomly moving and replacing initial positions with new positions; The individual goats gradually move toward the current optimal solution and continue to update their positions; The goat escapes from the local optimal solution through a jumping mechanism; For goat individuals whose fitness ranking is lower than the preset value, a new position is randomly generated for them; Continue to calculate the fitness value according to the new position and update the current optimal solution; When the variance of the optimal solution in the goat population does not change significantly, the position update is stopped, and the final optimal solution generated at the time of stopping is used as the optimal hyperparameter of the XGBoost model.

8. A method for determining the operating efficiency of a feedwater pump turbine according to claim 1, characterized in that: Also includes: Comparing the actual operating efficiency with the predicted operating efficiency, evaluating the efficiency prediction model, and outputting an evaluation result; According to the evaluation result, the efficiency prediction model is adaptively adjusted.

9. A method for determining the operating efficiency of a feedwater pump turbine according to claim 8, characterized in that: The method of adaptively adjusting the efficiency prediction model includes: changing the number of features in the low-dimensional feature matrix to improve the performance of the efficiency prediction model; adjusting the parameters of the efficiency prediction model to improve the efficiency of the efficiency prediction model.

10. A feedwater pump steam turbine operating efficiency monitoring system, which adopts a feedwater pump steam turbine operating efficiency determination method according to any one of claims 1 to 6, characterized in that: include: Sensors for real-time monitoring of various operating parameters of the feedwater pump turbine, including steam parameters and water pump parameters; A data storage unit, used for storing the data transmitted by the sensor and converting it into a digital signal; A central processing unit, used for processing and analyzing the data in the data storage unit, and calculating the operating efficiency of the steam turbine; Intelligent control terminal, used to predict the operating efficiency of feedwater pump turbine.

Images