A gas power generation data analysis system

Abstract

The application discloses a kind of gas power generation data analysis systems, specifically related to data analysis field, including data acquisition module, data processing module, coupling analysis module, analysis evaluation module and operation optimization strategy module;Data acquisition module: utilize sensor array in gas turbine operation process real-time acquisition multidimensional parameter, provide basic data support for subsequent analysis;Data processing module: the multidimensional parameter of collection is adaptively filtered and handled, to improve data accuracy and reliability;Coupling analysis module: the time-frequency domain feature extraction of preprocessed data is carried out, and the coupling relationship of operating parameter and combustion characteristics are obtained;The application acquires gas turbine operating parameter by multidimensional sensor array in real time, combines adaptive filtering and improved time-frequency analysis algorithm, realizes the high-precision processing and feature extraction of operating data, compared with traditional method, temperature data noise suppression rate is improved.

Description

Technical Field

[0001] This invention relates to the field of data analysis technology, and more specifically, to a gas-fired power generation data analysis system. Background Technology

[0002] In the field of gas-fired power generation, accurate analysis of equipment operating status is crucial for power generation efficiency and safety.

[0003] Traditional data analysis methods often rely on single sensor data or simple threshold judgments, making it difficult to address the multi-dimensional parameter coupling effects under complex operating conditions of gas turbines. For example, existing solutions often suffer from delayed fault warnings due to insufficient data fusion dimensions when analyzing the correlation between gas temperature, pressure fluctuations, and vibration signals. In terms of combustion efficiency optimization, adjusting gas flow rate based on a single parameter can easily lead to combustion instability. Furthermore, traditional methods predict equipment degradation trends based on historical data averages, failing to fully consider the time-varying characteristics of real-time operating data, resulting in significant prediction errors. When early faults such as minor turbine blade wear or combustion chamber carbon buildup occur, traditional methods, due to insufficient accuracy in feature extraction algorithms, struggle to achieve early fault identification, potentially triggering cascading failures and increasing maintenance costs and downtime losses.

[0004] Therefore, there is an urgent need for a gas-fired power generation data analysis system that can collect key operating parameters by deploying a multi-dimensional sensor network, perform data preprocessing using an adaptive filtering algorithm, analyze parameter coupling relationships by combining time-frequency domain feature extraction technology, and construct a hybrid fault diagnosis model based on physical models and data-driven approaches to achieve accurate assessment of equipment operating status and early warning of anomalies. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art, the present invention provides a gas-fired power generation data analysis system, which solves the problems mentioned in the background art through the following solutions.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a gas-fired power generation data analysis system, comprising:

[0007] Data acquisition module: Utilizes a sensor array to collect multi-dimensional parameters in real time during gas turbine operation, providing basic data support for subsequent analysis;

[0008] Data processing module: Performs adaptive filtering on the collected multidimensional parameters to improve data accuracy and reliability;

[0009] Coupling analysis module: Extracts time-frequency domain features from the preprocessed data to obtain the coupling relationship between operating parameters and combustion characteristics;

[0010] Analysis and evaluation module: Constructs a fault diagnosis model that integrates physical models and data-driven approaches to achieve accurate assessment of operational status;

[0011] Operation optimization strategy module: Based on fault diagnosis results and status assessment, implement operation parameter optimization to improve power generation efficiency and equipment reliability.

[0012] Preferably, the multidimensional parameters include combustion gas temperature, compressor outlet pressure, rotor vibration acceleration, fuel flow rate, and exhaust gas composition; the combustion gas temperature is acquired by an infrared temperature sensor located at the combustion chamber outlet; the compressor outlet pressure is acquired by a high-precision piezoresistive pressure sensor; the rotor vibration acceleration is acquired by a triaxial acceleration sensor; the fuel flow rate is acquired using a Coriolis mass flow meter; and the exhaust gas composition is acquired by an infrared spectrometer gas analyzer, with real-time monitoring of CO and [other components]. Concentration; simultaneously, during the data acquisition process, the percentage of equipment operating load L is recorded synchronously.

[0013] Preferably, the adaptive filtering process includes temperature preprocessing and pressure preprocessing; the temperature preprocessing: for the temperature sequence T(t), a dynamic threshold filtering algorithm based on a sliding window is adopted, specifically expressed as: ,in, This represents a sliding window centered at t, with the window length set to the average temperature over 100 sampling points. The standard deviation of the temperature within the window is represented by k, which represents the adaptive adjustment coefficient. k is determined by linear interpolation based on the current equipment operating load percentage L: k = 1.5 + 0.01 × L. The pressure preprocessing involves wavelet transform denoising of the pressure data P(t), using a db4 wavelet for 3-level decomposition, and applying soft thresholding to the high-frequency coefficients. The threshold function is: Where m represents the index of the high-frequency coefficients in the j-th layer, and n represents the number of sampling points. This represents the median of the high-frequency coefficients in the j-th layer.

[0014] Preferably, the time-frequency domain feature extraction includes vibration signal time-frequency feature extraction, temperature-pressure coupling feature extraction, and combustion feature parameter extraction; the vibration signal time-frequency feature extraction employs an improved short-time Fourier transform (STFT), selects a Hanning window as the window function, sets the window length to 256 sampling points, and the overlap rate to 75%, and calculates the energy entropy of the time-frequency matrix. ,in, Represents the elements of the time-frequency matrix after STFT transformation. The total energy of the time-frequency matrix is ​​represented by M and N, which represent the number of rows and columns of the time-frequency matrix, respectively; the temperature-pressure coupling feature extraction is defined by the temperature-pressure fluctuation coefficient. ,in, These represent the normalized fluctuation sequences of temperature and pressure, respectively. ) represents the covariance between the two. and Each represents its own standard deviation; the combustion characteristic parameter extraction: calculating fuel flow rate and exhaust gas... Dynamic ratio of concentration ,in, Indicates fuel flow rate. express concentration, The value represents the temperature influence coefficient, and L represents the percentage of equipment operating load.

[0015] Preferably, the fault diagnosis model includes the construction of a normal operating condition feature library, an anomaly identification algorithm, and a fault location method; the construction of the normal operating condition feature library involves collecting stable operating data of the equipment under different loads, namely 25%, 50%, 75%, and 100%, calculating the normal range boundary values ​​of each feature parameter, and establishing a probability distribution model of the feature parameters using the kernel density estimation (KDE) method; the anomaly identification algorithm involves calculating the Mahalanobis distance between the real-time feature parameter X and the normal operating condition feature library. ,in, This represents the mean vector of characteristics under normal operating conditions. Represents the covariance matrix. The sign for the transpose of a matrix, and when When the threshold is exceeded, an anomaly is determined; the fault location method: when an anomaly is detected, key influencing parameters are determined through sensitivity analysis, and parameter sensitivity is defined. ,in, This represents the i-th feature parameter. Let represent the normal operating mean of the i-th feature parameter, and select . The top three largest parameters are used as key parameters for fault location, and the cause of the fault is inferred by combining them with the physical model of the equipment.

[0016] Preferably, the operating parameter optimization includes a dynamic fuel flow adjustment algorithm, load reduction early warning, and maintenance suggestions; the dynamic fuel flow adjustment algorithm: when combustion instability characteristics are identified, the dynamic fuel flow adjustment algorithm is activated, and the adjustment amount... ,in, This represents the adjustment coefficient, which is determined based on the current equipment operating load percentage L. =1.5+0.01×L; The load reduction warning: when the rotor vibration energy entropy When the load exceeds 1.5 times the normal threshold, a load reduction warning is triggered, and the load adjustment rate... ,in, This indicates the normal threshold. This indicates the maximum load; the maintenance recommendations are as follows: Based on the equipment's degradation trend, long-term trends in vibration and temperature characteristics are used to determine maintenance recommendations, including turbine blade cleaning cycle recommendations. ,in, This represents the integral variable, which is any instantaneous operating moment within the cumulative operating time t of the gas turbine from startup. Indicates the baseline cleaning cycle. This represents the degradation impact coefficient.

[0017] The technical effects and advantages of this invention are as follows:

[0018] 1. This invention acquires gas turbine operating parameters in real time through a multi-dimensional sensor array, and combines adaptive filtering and improved time-frequency analysis algorithms to achieve high-precision processing and feature extraction of operating data. Compared with traditional methods, the noise suppression rate of temperature data is improved, the fault feature identification accuracy of vibration signals is increased, and the analysis error caused by data interference under complex operating conditions is effectively solved.

[0019] 2. This invention integrates physical models and data-driven algorithms through a hybrid fault diagnosis model, and achieves rapid fault location through Mahalanobis distance and parameter sensitivity analysis; it provides early warning of typical faults such as combustion instability and bearing wear, avoids further expansion of faults, and reduces operation and maintenance costs;

[0020] 3. This invention, through a feature-based decoupling-based operation optimization strategy, can dynamically adjust operating parameters based on real-time analysis results; thus reducing... Emissions are reduced, fuel efficiency is improved, and the number of unplanned equipment downtimes is reduced, significantly improving the economy and reliability of gas-fired power generation systems. Attached Figure Description

[0021] Figure 1 This is a schematic diagram of the overall structure of the present invention. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. 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 are within the scope of protection of the present invention.

[0023] As attached Figure 1 The gas-fired power generation data analysis system shown includes a data acquisition module, a data processing module, a coupled analysis module, an analysis and evaluation module, and an operation optimization strategy module.

[0024] The data acquisition module uses a sensor array to collect multi-dimensional parameters in real time during the operation of the gas turbine, providing basic data support for subsequent analysis.

[0025] It should be noted that the multidimensional parameters include gas temperature, compressor outlet pressure, rotor vibration acceleration, fuel flow rate, and exhaust composition; the gas temperature is acquired by an infrared temperature sensor located at the combustion chamber outlet; the compressor outlet pressure is acquired using a high-precision piezoresistive pressure sensor; the rotor vibration acceleration is acquired by a triaxial acceleration sensor; the fuel flow rate is acquired using a Coriolis mass flow meter; and the exhaust composition is acquired by an infrared spectrometer gas analyzer, with real-time monitoring of CO and [other components]. Concentration; simultaneously, during the data acquisition process, the percentage of equipment operating load L is recorded synchronously.

[0026] The data processing module performs adaptive filtering on the collected multidimensional parameters to improve data accuracy and reliability.

[0027] It should be noted that the adaptive filtering process includes temperature preprocessing and pressure preprocessing; the temperature preprocessing, for the temperature sequence T(t), employs a dynamic threshold filtering algorithm based on a sliding window, specifically expressed as follows: ,in, This represents a sliding window centered at t, with the window length set to the average temperature over 100 sampling points. The standard deviation of the temperature within the window is represented by k, which represents the adaptive adjustment coefficient. k is determined by linear interpolation based on the current equipment operating load percentage L: k = 1.5 + 0.01 × L. The pressure preprocessing involves wavelet transform denoising of the pressure data P(t), using a db4 wavelet for 3-level decomposition, and applying soft thresholding to the high-frequency coefficients. The threshold function is: Where m represents the index of the high-frequency coefficients in the j-th layer, and n represents the number of sampling points. This represents the median of the high-frequency coefficients in the j-th layer.

[0028] The coupling analysis module extracts time-frequency domain features from the preprocessed data to obtain the coupling relationship between operating parameters and combustion characteristics.

[0029] It should be noted that the time-frequency domain feature extraction includes vibration signal time-frequency feature extraction, temperature-pressure coupling feature extraction, and combustion feature parameter extraction; the vibration signal time-frequency feature extraction uses an improved short-time Fourier transform (STFT), with a Hanning window as the window function, a window length of 256 sampling points, and an overlap rate of 75%, and calculates the energy entropy of the time-frequency matrix. ,in, Represents the elements of the time-frequency matrix after STFT transformation. The total energy of the time-frequency matrix is ​​represented by M and N, which represent the number of rows and columns of the time-frequency matrix, respectively; the temperature-pressure coupling feature extraction is defined by the temperature-pressure fluctuation coefficient. ,in, These represent the normalized fluctuation sequences of temperature and pressure, respectively. ) represents the covariance between the two. and Each represents its own standard deviation; the combustion characteristic parameter extraction: calculating fuel flow rate and exhaust gas... Dynamic ratio of concentration ,in, Indicates fuel flow rate. express concentration, The value represents the temperature influence coefficient, and L represents the percentage of equipment operating load.

[0030] The analysis and evaluation module: constructs a fault diagnosis model that integrates physical models and data-driven approaches, thereby achieving accurate assessment of operational status;

[0031] It should be noted that the fault diagnosis model includes the construction of a normal operating condition feature library, an anomaly identification algorithm, and a fault location method. The construction of the normal operating condition feature library involves collecting stable operating data of the equipment under different loads (25%, 50%, 75%, and 100%), calculating the normal range boundary values ​​of each feature parameter, and establishing a probability distribution model of the feature parameters using the kernel density estimation (KDE) method. The anomaly identification algorithm calculates the Mahalanobis distance between the real-time feature parameter X and the normal operating condition feature library. ,in, This represents the mean vector of characteristics under normal operating conditions. Represents the covariance matrix. The sign for the transpose of a matrix, and when When the threshold is exceeded, an anomaly is determined; the fault location method: when an anomaly is detected, key influencing parameters are determined through sensitivity analysis, and parameter sensitivity is defined. ,in, This represents the i-th feature parameter. Let represent the normal operating mean of the i-th feature parameter, and select . The top three largest parameters are used as key parameters for fault location, and the cause of the fault is inferred by combining them with the physical model of the equipment.

[0032] The operation optimization strategy module optimizes operating parameters based on fault diagnosis results and status assessment to improve power generation efficiency and equipment reliability.

[0033] It should be noted that the aforementioned operating parameter optimization includes a dynamic fuel flow adjustment algorithm, load reduction early warning, and maintenance recommendations. The dynamic fuel flow adjustment algorithm is activated when combustion instability characteristics are identified, adjusting the flow rate accordingly. ,in, This represents the adjustment coefficient, which is determined based on the current equipment operating load percentage L. =1.5+0.01×L; The load reduction warning: when the rotor vibration energy entropy When the load exceeds 1.5 times the normal threshold, a load reduction warning is triggered, and the load adjustment rate... ,in, This indicates the normal threshold. This indicates the maximum load; the maintenance recommendations are as follows: Based on the equipment's degradation trend, long-term trends in vibration and temperature characteristics are used to determine maintenance recommendations, including turbine blade cleaning cycle recommendations. ,in, This represents the integral variable, which is any instantaneous operating moment within the cumulative operating time t of the gas turbine from startup. Indicates the baseline cleaning cycle. This represents the degradation impact coefficient.

[0034] This invention acquires gas turbine operating parameters in real time using a multi-dimensional sensor array, and combines adaptive filtering with an improved time-frequency analysis algorithm to achieve high-precision processing and feature extraction of operating data. Compared with traditional methods, it improves the noise suppression rate of temperature data and the accuracy of fault feature identification of vibration signals, effectively solving the analysis error problem caused by data interference under complex operating conditions. This invention integrates a physical model and a data-driven algorithm through a hybrid fault diagnosis model, and achieves rapid fault location through Mahalanobis distance and parameter sensitivity analysis. It provides earlier warnings for typical faults such as combustion instability and bearing wear, preventing further fault expansion and reducing operation and maintenance costs. This invention uses a feature-based decoupling-based operation optimization strategy to dynamically adjust operating parameters based on real-time analysis results, reducing... Emissions are reduced, fuel efficiency is improved, and the number of unplanned equipment downtimes is reduced, significantly improving the economy and reliability of gas-fired power generation systems.

[0035] Secondly: The accompanying drawings of the embodiments disclosed in this invention only involve the structures involved in the embodiments disclosed in this invention. Other structures can refer to the general design. In the absence of conflict, the same embodiment and different embodiments of this invention can be combined with each other.

[0036] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A gas-fired power generation data analysis system, characterized in that, include: Data acquisition module: Utilizes a sensor array to collect multi-dimensional parameters in real time during gas turbine operation, providing basic data support for subsequent analysis; Data processing module: Performs adaptive filtering on the collected multidimensional parameters to improve data accuracy and reliability; Coupling analysis module: Extracts time-frequency domain features from the preprocessed data to obtain the coupling relationship between operating parameters and combustion characteristics; The time-frequency domain feature extraction includes vibration signal time-frequency feature extraction, temperature-pressure coupling feature extraction, and combustion feature parameter extraction. The vibration signal time-frequency feature extraction employs an improved Short-Time Fourier Transform (STFT), using a Hanning window as the window function, with a window length of 256 sampling points and an overlap rate of 75%, and calculates the energy entropy of the time-frequency matrix. ,in, Represents the elements of the time-frequency matrix after STFT transformation. The total energy of the time-frequency matrix is ​​represented by M and N, which represent the number of rows and columns of the time-frequency matrix, respectively; the temperature-pressure coupling feature extraction is defined by the temperature-pressure fluctuation coefficient. ,in, These represent the normalized fluctuation sequences of temperature and pressure, respectively. ) represents the covariance between the two. and Each represents its own standard deviation; the combustion characteristic parameter extraction: calculating fuel flow rate and exhaust gas... Dynamic ratio of concentration ,in, Indicates fuel flow rate. express concentration, The coefficient of performance is represented by temperature, and L represents the percentage of equipment operating load. Analysis and evaluation module: Constructs a fault diagnosis model that integrates physical models and data-driven approaches to achieve accurate assessment of operational status; Operation optimization strategy module: Based on fault diagnosis results and status assessment, implement operational parameter optimization to improve power generation efficiency and equipment reliability; The operational parameter optimization includes a dynamic fuel flow adjustment algorithm, load reduction early warning, and maintenance recommendations. The dynamic fuel flow adjustment algorithm is activated when combustion instability characteristics are identified, adjusting the flow rate accordingly. ,in, This represents the adjustment coefficient, determined based on the current equipment operating load percentage L. =1.5+0.01×L; The load reduction warning: when the rotor vibration energy entropy When the load exceeds 1.5 times the normal threshold, a load reduction warning is triggered, and the load adjustment rate... ,in, Indicates the normal threshold. This indicates the maximum load; the maintenance recommendations are as follows: Based on the equipment's deterioration trend, long-term changes in vibration and temperature characteristics are used to determine maintenance recommendations, including turbine blade cleaning cycle recommendations. ,in, This represents the integral variable, which is any instantaneous operating moment within the cumulative operating time t of the gas turbine from startup. Indicates the baseline cleaning cycle. This represents the degradation impact coefficient.

2. The gas-fired power generation data analysis system according to claim 1, characterized in that: The multidimensional parameters include gas temperature, compressor outlet pressure, rotor vibration acceleration, fuel flow rate, and exhaust composition. The gas temperature is acquired using an infrared temperature sensor located at the combustion chamber outlet. The compressor outlet pressure is acquired using a high-precision piezoresistive pressure sensor. The rotor vibration acceleration is acquired using a triaxial acceleration sensor. The fuel flow rate is acquired using a Coriolis mass flow meter. The exhaust composition is acquired using an infrared gas spectrometer, with real-time monitoring of CO and other components. Concentration; simultaneously, during the data acquisition process, the percentage of equipment operating load L is recorded synchronously.

3. The gas-fired power generation data analysis system according to claim 1, characterized in that: The adaptive filtering process includes temperature preprocessing and pressure preprocessing; the temperature preprocessing: for the temperature sequence T(t), a dynamic threshold filtering algorithm based on a sliding window is adopted, specifically expressed as: ,in, This represents a sliding window centered at t, with the window length set to the average temperature over 100 sampling points. The standard deviation of the temperature within the window is represented by k, which represents the adaptive adjustment coefficient. k is determined by linear interpolation based on the current equipment operating load percentage L: k = 1.5 + 0.01 × L. The pressure preprocessing involves wavelet transform denoising of the pressure data P(t), using a db4 wavelet for 3-level decomposition, and applying soft thresholding to the high-frequency coefficients. The threshold function is: Where m represents the index of the high-frequency coefficients in the j-th layer, and n represents the number of sampling points. This represents the median of the high-frequency coefficients in the j-th layer.

4. The gas-fired power generation data analysis system according to claim 1, characterized in that: The fault diagnosis model includes the construction of a normal operating condition feature library, an anomaly identification algorithm, and a fault location method. The normal operating condition feature library construction involves collecting stable operating data of the equipment under different loads (25%, 50%, 75%, and 100%), calculating the normal range boundary values ​​of each feature parameter, and establishing a probability distribution model of the feature parameters using the kernel density estimation (KDE) method. The anomaly identification algorithm calculates the Mahalanobis distance between the real-time feature parameter X and the normal operating condition feature library. ,in, This represents the mean vector of characteristics under normal operating conditions. Represents the covariance matrix. The sign for the transpose of a matrix, and when When the threshold is exceeded, an anomaly is determined; the fault location method: when an anomaly is detected, key influencing parameters are determined through sensitivity analysis, and parameter sensitivity is defined. ,in, This represents the i-th feature parameter. Let represent the normal operating mean of the i-th feature parameter, and select . The top three largest parameters are used as key parameters for fault location, and the cause of the fault is inferred by combining them with the physical model of the equipment.

Images