Gas turbine power plant generating capacity prediction system and method thereof

Abstract

The invention provides a gas turbine power plant generating capacity prediction method comprising the following steps: obtaining historical generating capacity data including historical generating capacity data and generating equipment operation data; analyzing the data; preprocessing the data, including data cleaning, time feature extraction, power generation equipment operation data processing and feature selection; selecting a time sequence prediction model for training and prediction; evaluating the prediction performance of the model, and adjusting and optimizing the model; methods such as cross validation, adjustment of model parameters or integration of multiple models may be used. The invention discloses a gas turbine power plant generating capacity prediction system, which comprises a data acquisition module, a model training module, a prediction module and an equipment operation data acquisition module, and is characterized in that historical generating capacity and equipment data are utilized, a time sequence prediction model and an algorithm are combined, and a prediction model is trained, evaluated and trained; therefore, planning and operation of the power plant are facilitated.

Description

Technical Field

[0001] The present invention belongs to the technical field of plant entry safety management, and specifically refers to a system and method for predicting power generation in a gas turbine power plant. Background Art

[0002] Gas turbine power plant is the abbreviation of gas turbine power plant, which is a manifestation of clean power generation technology. The fuel of gas turbine power plant is natural gas, fuel oil or industrial associated gas, etc. Due to the complete combustion, the emissions generated by its combustion have little impact on the environment and little noise pollution; and because gas turbine power plant has the advantages of fast installation, small size, low investment, high efficiency, low operating cost and long life cycle, it is currently widely used in the market.

[0003] Accurate prediction of power generation is the basis for the healthy operation and development of power plants. Inaccurate prediction of power generation will lead to regional disadvantages and energy waste. In addition, the mismatch between peak power generation and peak power consumption has become the core contradiction between power generation and power consumption.

[0004] Based on historical data, using appropriate models and data mining technology to predict future power generation can provide a reference for power plants to formulate dispatch plans. At present, developed countries in Europe and the United States have begun to use power generation forecasting as an important reference for power grid dispatching. Power generation forecasting has become an important part of the control and management of large-scale power plant systems.

[0005] The current power generation forecasting system has not kept pace with the development of the big data era and has certain limitations in terms of time series information. It is unable to mine the effective information and potential relationships contained in non-continuous data, is unable to fully capture the correlation between data and related influencing factors, and is not flexible enough in extracting features from non-continuous data. Summary of the invention

[0006] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides a system and method for predicting power generation of a gas turbine power plant.

[0007] In one aspect, the present invention provides a method for predicting power generation of a gas turbine power plant, comprising:

[0008] Obtain historical power generation data, including historical power generation data and power generation equipment operation data; analyze the data;

[0009] Preprocess the data, including data cleaning, extraction of time features, power generation equipment operation data processing, and feature selection;

[0010] Select a time series forecasting model for training and forecasting;

[0011] Evaluate the performance of model predictions and tune the model; you can use methods such as cross-validation, adjusting model parameters, or integrating multiple models.

[0012] The trained model can be used to predict future power generation and combined with the operating data of actual power generation equipment to obtain more accurate prediction results.

[0013] Furthermore, the historical power generation data includes time series data, including daily, weekly, monthly and annual power generation data.

[0014] Furthermore, the power generation equipment operation data includes data on factors that affect the power generation of the equipment, including the type and form of the equipment, the configuration of the equipment, the operating temperature of the equipment, the gas path, the intake temperature, and the humidity.

[0015] Furthermore, the data analysis includes the time range and frequency of the data, whether the data is missing or abnormal; data distribution and data trend, and the correlation between power generation data and power generation equipment operation data.

[0016] Furthermore, the data cleaning includes processing missing data and abnormal data; extracting time features, extracting relevant information of the day, week, month and year from the timestamp; processing of power generation equipment operation data includes standardization processing to ensure comparability between data of different metrics; feature selection includes selecting features related to power generation.

[0017] Furthermore, the time series prediction model includes a linear regression model, a decision tree model, a support vector machine model, and a neural network model;

[0018] The linear regression model is used for data with a high linear relationship. In the linear regression model, power generation is used as the dependent variable, and the operating status characteristics of the power generation equipment are used as the independent variable. The operating data of various indicators of the power generation equipment that affect the power generation are collected. The data is collected through sensors and detection systems, and the collected data is screened. Through correlation analysis, independent variables with a high correlation with power generation are found and incorporated into the model. The regression equation coefficients are obtained through the least squares method. Data are collected through statistical software for regression analysis to obtain the regression equation. The prediction accuracy of the model is evaluated by verifying the goodness of fit.

[0019] A decision tree model is used for nonlinear, non-continuous, and missing value data. In the decision tree model, the equipment operation data is collected, and the XGBoost algorithm is used to train and generate a gradient boosting decision tree model. The equipment operation data is used to obtain the power generation data through the decision tree model, with the data of the power generation device as the input value and the power generation as the output value;

[0020] For the processing of missing values, select the optimal partition attribute

[0021]

[0022] in,

[0023]

[0024] ρ is the proportion of samples without missing values, is the proportion of the kth class in the sample without missing values, Take the value a for attribute a in the sample with no missing values v The sample proportion of

[0025] The support vector machine model is used for data with small sample size and high latitude. First, the power generation equipment information is used to obtain the power generation of the gas turbine. The input content includes historical data and power generation equipment information data, and the output content is the power generation and power generation.

[0026] The inlet temperature and humidity sequence of the gas turbine is decomposed into n modal components of different scales by using variational mode decomposition. The temperature and humidity sequences corresponding to the n modal components are used to establish LSSVM prediction models of the least squares support vector machine. The data sets of the temperature and humidity sequences are divided into training set, cross-validation set and test set, and the data are normalized. The LSSVM prediction model is trained by rolling the training set. The particle swarm optimization algorithm is used to optimize the kernel width and penalty factor of the LSSVN prediction model on the cross-validation set. Then the error is tested on the test set to obtain the optimal LSSVM prediction model with an error threshold less than the error threshold.

[0027] The neural network model is used for complex relationships and large-scale data. The input content is the power generation equipment information data, and the output content is the power generation data. The error after the output is used to estimate the error of the direct predecessor layer of the output layer, and the error is used to estimate the error of the previous layer. The activation function is the s function, f(x) = 1 / (1+ex). If the actual output of the output layer does not match the expected output, the error is apportioned to all units in each layer, thereby correcting the weights of the units in each layer. When the output error reaches an acceptable level, the training ends.

[0028] Furthermore, the prediction performance of the model was evaluated, and the evaluation indicators included: mean square error, root mean square error, and mean absolute error.

[0029] On the other hand, the present invention discloses a power generation prediction system for a gas turbine power plant, comprising a data acquisition module, a model training module, a prediction module, and an equipment operation data acquisition module, wherein the data acquisition module acquires historical power generation data, including time series data, including daily, weekly, monthly, and annual power generation data; the equipment operation data acquisition module includes a sensor and a detection system, which collects data on factors affecting the power generation of the equipment, including the type and form of the equipment, the configuration of the equipment, the operating temperature of the equipment, the gas path, the intake temperature, and the humidity; the model training module trains and optimizes the historical data based on the historical power generation data and the equipment operation data; after being trained, the prediction module combines the operation data of the actual power generation equipment to predict the power generation in the future.

[0030] The beneficial effects achieved by the present invention using the above structure are as follows: This scheme provides a gas turbine power plant power generation prediction system and method, which targets the characteristics of the gas turbine, the correlation between the power generation power and the operating temperature, gas path, intake temperature, and humidity of the equipment, utilizes historical power generation and equipment data, and combines time series prediction models and algorithms to train and evaluate the prediction models, accurately predict power generation, and provide precise data for the company's production and scheduling, which is beneficial to the planning and operation of the power plant, improves energy utilization, and achieves the effect of reducing costs and increasing efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 It is a flow chart of a method for predicting power generation of a gas turbine power plant;

[0032] Figure 2 The figure is a framework diagram of a power generation prediction system for a gas turbine power plant. DETAILED DESCRIPTION

[0033] The technical solutions in the embodiments of the present invention are clearly and completely described below. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0034] The present invention provides a method for predicting power generation of a gas turbine power plant, comprising:

[0035] S1. Obtain historical power generation data, including historical power generation data and time series data of historical power generation, including daily, weekly, monthly and annual power generation;

[0036] Power generation equipment operation data and data related to power generation, including equipment type and form, equipment configuration, equipment operating temperature, gas path, inlet temperature, humidity

[0037] Analyze the time range and frequency of the data, analyze whether there are any missing data, and analyze whether there are any anomalies in the data;

[0038] Analyze the distribution and trend of data, and analyze the correlation between equipment operation data and power generation.

[0039] S2. Preprocess the data, including data cleaning, and process missing and abnormal data;

[0040] Extraction of time features: extracting relevant information of day, week, month and year from timestamp;

[0041] Processing of power generation equipment operation data, normalizing equipment information to make data from different units comparable;

[0042] Feature extraction, based on correlation analysis, select features related to power generation;

[0043] S3, model selection, select time series prediction models for training and prediction, including linear regression model, decision tree model, support vector machine model, neural network model;

[0044] S4. Evaluate the predictive performance of the model, including mean square error, root mean square error, mean absolute error, use cross-validation, adjust model parameters, and integrate multiple models;

[0045] S5. Use the trained model to predict future power generation and combine it with the operating data of actual power generation equipment to obtain more accurate prediction results.

[0046] The linear regression model is used for data with a high linear relationship. In the linear regression model, power generation is used as the dependent variable, and the operating status characteristics of the power generation equipment are used as the independent variable. The operating data of various indicators of the power generation equipment that affect the power generation are collected. The data is collected through sensors and detection systems, and the collected data is screened. Through correlation analysis, the independent variables with a high correlation with power generation are found and incorporated into the model. The regression equation coefficients are obtained through the least squares method. Data is collected through statistical software for regression analysis to obtain the regression equation. The prediction accuracy of the model is evaluated by verifying the goodness of fit.

[0047] A decision tree model is used for nonlinear, non-continuous, and missing value data. In the decision tree model, the equipment operation data is collected, and the XGBoost algorithm is used to train and generate a gradient boosting decision tree model. The equipment operation data is used to obtain the power generation data through the decision tree model, with the data of the power generation device as the input value and the power generation as the output value;

[0048] For the processing of missing values, select the optimal partition attribute

[0049]

[0050] in,

[0051]

[0052] ρ is the proportion of samples without missing values, is the proportion of the kth class in the sample without missing values, Take the value a for attribute a in the sample with no missing values v The sample proportion of

[0053] The support vector machine model is used for data with small sample size and high latitude. First, the power generation equipment information is used to obtain the power generation of the gas turbine. The input content includes historical data and power generation equipment information data, and the output content is the power generation and power generation.

[0054] The inlet temperature and humidity sequence of the gas turbine is decomposed into n modal components of different scales by using variational mode decomposition. The temperature and humidity sequences corresponding to the n modal components are used to establish LSSVM prediction models of the least squares support vector machine. The data sets of the temperature and humidity sequences are divided into training set, cross-validation set and test set, and the data are normalized. The LSSVM prediction model is trained by rolling the training set. The particle swarm optimization algorithm is used to optimize the kernel width and penalty factor of the LSSVN prediction model on the cross-validation set. Then the error is tested on the test set to obtain the optimal LSSVM prediction model with an error threshold less than the error threshold.

[0055] The neural network model is used for complex relationships and large-scale data. The input content is the power generation equipment information data, and the output content is the power generation data. The error after the output is used to estimate the error of the direct predecessor layer of the output layer, and the error is used to estimate the error of the previous layer. The activation function is the s function, f(x) = 1 / (1+ex). If the actual output of the output layer does not match the expected output, the error is apportioned to all units in each layer, thereby correcting the weights of the units in each layer. When the output error reaches an acceptable level, the training ends.

[0056] In an embodiment, one year of historical data of the power plant is obtained, including power generation data and equipment operation data, and the historical data and the correlation between the inlet temperature and humidity of the gas turbine and the power generation are analyzed. In the feature extraction stage, the time features of the day, week, month and year are extracted, the inlet temperature and humidity of the gas turbine are normalized, and the autocorrelation and partial autocorrelation diagrams are used to determine the appropriate lag value.

[0057] Correlation is the degree of correlation between two variables. Assuming that each variable conforms to the normal distribution, the Pearson correlation coefficient is used to calculate the correlation between variables. The Pearson correlation coefficient is a number between -1 and 1, describing negative or positive correlation respectively. A value of zero indicates no correlation.

[0058] Select the linear regression model, train and optimize the historical data within one year, and obtain the final model through cross-validation and adjustment of model parameters.

[0059] The specific steps of cross validation are: randomly divide the training data into n parts, s1, s2, ...sn;

[0060] Perform the algorithm n times for each model Mi, each time select one sj as the validation set, and the others as the training set to train the model Mi, and test the trained model Mi on sj. Each test gets an error e, and n errors are obtained. The n errors are averaged to get the generalization error.

[0061] Select the model with the smallest generalization error and train it again on the training set to obtain the final model.

[0062] The number of training samples is n=10.

[0063] When performing feature selection, the training data is randomly divided into n parts, s1, s2, ...sn;

[0064] The algorithm performs feature selection on each validation data sj, constructs a text classifier, and calculates the generalization error on the validation set sj;

[0065] Take the average of the n errors to get the generalization error;

[0066] Select the model with the smallest generalization error and train it again on the training set to obtain the final model.

[0067] The present invention discloses a power generation prediction system for a gas turbine power plant, comprising a data acquisition module, a model training module, a prediction module, and an equipment operation data acquisition module, wherein the data acquisition module acquires historical power generation data, including time series data, including daily, weekly, monthly, and annual power generation data; the equipment operation data acquisition module comprises a sensor and a detection system, which collects data on factors affecting the power generation of the equipment, including the type and form of the equipment, the configuration of the equipment, the operating temperature of the equipment, the gas path, the intake temperature, and the humidity; the model training module trains and optimizes the historical data within one year based on the historical power generation data and the equipment operation data; after being trained, the prediction module predicts the power generation within a period of time in the future in combination with the operation data of the actual power generation equipment.

[0068] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

[0069] The present invention and its embodiments are described above, and such description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if ordinary technicians in the field are inspired by it, without departing from the purpose of the invention, they can design a structure and embodiment similar to the technical solution without creativity, which should belong to the protection scope of the present invention.

Claims

1. A method for predicting power generation of a gas turbine power plant, characterized in that: include: Obtain historical power generation data, including historical power generation data and power generation equipment operation data; analyze the data; Preprocess the data, including data cleaning, extraction of time features, power generation equipment operation data processing, and feature selection; Select a time series forecasting model for training and forecasting; Evaluate the performance of model predictions and tune the model; you can use methods such as cross-validation, adjusting model parameters, or integrating multiple models. The trained model can be used to predict future power generation and combined with the operating data of actual power generation equipment to obtain more accurate prediction results.

2. A method for predicting power generation of a gas turbine power plant according to claim 1, characterized in that: The historical power generation data includes time series data, including daily, weekly, monthly and annual power generation data.

3. A method for predicting power generation of a gas turbine power plant according to claim 1, characterized in that: The power generation equipment operation data includes data on factors that affect the power generation of the equipment, including the type and form of the equipment, the configuration of the equipment, the operating temperature of the equipment, the gas path, the intake temperature, and the humidity.

4. A method for predicting power generation of a gas turbine power plant according to claim 1, characterized in that: The data analysis includes the time range and frequency of the data, whether the data is missing or abnormal; the data distribution and data trend, and the correlation between the power generation data and the power generation equipment operation data.

5. A method for predicting power generation of a gas turbine power plant according to claim 1, characterized in that: The data cleaning includes processing missing data and abnormal data; extracting time features, extracting relevant information of day, week, month and year from timestamps; processing of power generation equipment operation data includes standardization processing to ensure comparability between data of different metrics; feature selection includes selecting features related to power generation.

6. A method for predicting power generation of a gas turbine power plant according to claim 1, characterized in that: The time series prediction model includes a linear regression model, a decision tree model, a support vector machine model, and a neural network model.

7. A method for predicting power generation of a gas turbine power plant according to claim 1, characterized in that: Evaluate the prediction performance of the model. The evaluation indicators include mean square error, root mean square error, and mean absolute error.

8. A method for predicting power generation of a gas turbine power plant according to claim 1, characterized in that: The steps of cross validation are to randomly divide the training data into n parts, s1, s2, ...sn; Perform the algorithm n times for each model Mi, each time select one sj as the validation set, and the others as the training set to train the model Mi, and test the trained model Mi on sj. Each test gets an error e, and n errors are obtained. The n errors are averaged to get the generalization error. Select the model with the smallest generalization error and train it again on the training set to obtain the final model; When performing feature selection, the training data is randomly divided into n parts, s1, s2, ...sn; The algorithm performs feature selection on each validation data sj, constructs a text classifier, and calculates the generalization error on the validation set sj; Take the average of the n errors to get the generalization error; Select the model with the smallest generalization error and train it again on the training set to obtain the final model.

9. A power generation prediction system for a gas turbine power plant, characterized in that: It includes a data acquisition module, a model training module, a prediction module, and an equipment operation data acquisition module. The data acquisition module acquires historical power generation data, including time series data, including daily, weekly, monthly, and annual power generation data; the equipment operation data acquisition module includes sensors and detection systems to collect data on factors that affect the power generation of the equipment, including the type and form of the equipment, the configuration of the equipment, the operating temperature of the equipment, the gas path, the intake temperature, and the humidity; the model training module trains and optimizes the historical data based on the historical power generation data and the equipment operation data; after training, the prediction module combines the operation data of the actual power generation equipment to predict the power generation in the future.

Images