Estimation method of steam turbine heat rate based on operation data

Abstract

The application discloses a steam turbine heat consumption estimation method based on operation data, and constructs a three-dimensional acquisition system of 'core parameters + auxiliary parameters + environmental parameters'. The core parameters are collected by a DCS system, the auxiliary parameters are obtained through newly-added special sensors, and the environmental parameters are collected by relying on a meteorological station module. Qualified input data are formed after the original data are processed through denoising, missing value completion and abnormal data elimination. The core calculation formula is derived based on the first law of thermodynamics, six types of loss quantification sub-models are constructed, the load interval is divided through working condition identification, the loss correction coefficient is optimized by using historical data and a random forest algorithm, the auxiliary parameters are input into the model to obtain the coefficient, the preliminary heat consumption value is obtained by correcting the core formula result, the method improves the heat consumption estimation precision through comprehensive data acquisition, accurate model construction and dynamic correction, has a heat consumption deviation tracing function, and is suitable for different types of steam turbines and operation scene requirements.

Description

Technical Field

[0001] This invention relates to the field of steam turbine technology, and more specifically to a method for estimating steam turbine heat consumption based on operating data. Background Technology

[0002] Steam turbines are core prime movers in energy production sectors such as thermal power generation, and their heat rate is a key indicator for measuring equipment energy efficiency, power plant operating economy, and low-carbon operation capabilities. Accurate estimation of steam turbine heat consumption is an important technical support for equipment operating status monitoring and fault diagnosis, and is of great significance for power plant energy-saving renovation, optimized scheduling, and carbon emission reduction.

[0003] Current industry-standard methods for estimating turbine heat consumption, while meeting basic accounting needs, are ill-suited for refined operational requirements. Data acquisition is limited, relying heavily on DCS for core parameters without incorporating auxiliary parameters. Furthermore, low sampling frequency leads to insufficient data completeness and representativeness, failing to provide reliable foundational data for heat consumption estimation. Additionally, the lack of environmental adaptability, with environmental parameters often fixed according to standard operating conditions and the absence of dynamic correlation mechanisms, prevents compensation for environmental changes on heat consumption and amplifies estimation errors. Summary of the Invention

[0004] To address this, the present invention provides a method for estimating the heat consumption of steam turbines based on operating data, thereby overcoming the problems of the prior art.

[0005] This invention is implemented by the following technical solution:

[0006] The method for estimating turbine heat consumption based on operating data includes the following steps:

[0007] Step 1: Construct a three-dimensional data acquisition system consisting of "core parameters + auxiliary parameters + environmental parameters". Core parameters are acquired through a distributed control system (DCS) with increased sampling frequency, covering main steam pressure, main steam temperature, exhaust pressure, and power generation. Auxiliary parameters are acquired through newly added dedicated sensors, covering shaft seal leakage flow, low-pressure cylinder exhaust humidity, and high-pressure heater terminal temperature difference. Environmental parameters are acquired through a weather station module, covering ambient temperature, atmospheric pressure, and relative humidity. The raw data is transmitted to edge computing nodes, where sensor noise is removed using a Kalman filter algorithm, missing data is completed using Lagrange interpolation, and abnormal data is removed after parameter threshold verification to obtain qualified input data.

[0008] Step 2: Derive the core calculation formula for turbine heat consumption based on the first law of thermodynamics The model is defined as follows: D1 is the main steam flow rate, h1 is the main steam enthalpy, D2 is the reheat steam flow rate, h2 is the reheat steam enthalpy, D3 is the exhaust steam flow rate, h3 is the exhaust steam enthalpy, and P is the power generation. Simultaneously, quantitative sub-models are constructed for steam flow friction loss, dynamic and static clearance leakage loss, cylinder heat dissipation loss, valve throttling loss, steam residual velocity loss, and interstage leakage loss. The turbine load is divided into three intervals—low load, medium load, and full load—through the operating condition identification module, with model parameters configured independently for each interval. Historical operating data covering each load interval and different ambient temperatures are used, and the random forest algorithm is employed to optimize the correction coefficients for each loss term, resulting in a random forest corrected model. Auxiliary parameters are input into this model, and the correction coefficients are output, correcting the core heat consumption calculation formula to obtain the preliminary heat consumption value.

[0009] Step 3: Establish a correlation model between ambient temperature, atmospheric pressure, condenser vacuum, and turbine internal efficiency through regression analysis to obtain the environmental impact coefficient; collect operating parameters such as turbine lubricating oil temperature, bearing vibration, and heater terminal temperature difference to determine the equipment operating status. Under normal conditions, the abnormal correction coefficient is set to 1; under abnormal conditions, the correction coefficient is set to "1". (ΔX is the deviation between the actual value and the standard value of the parameter, and X0 is the standard value of the parameter) Calculate the anomaly correction coefficient, obtain the comprehensive correction coefficient based on the two types of coefficients, substitute it into the preliminary heat consumption value to obtain the final heat consumption value, compare and verify the final heat consumption value with historical data, output if the deviation is ≤2%, and optimize the model parameters and recalculate if the deviation exceeds the standard. At the same time, accumulate the comparison results of actual operation and heat consumption test data, optimize the associated model parameters to realize the algorithm self-learning.

[0010] Preferably, the three-dimensional acquisition system in step 1 needs to cover the entire steam flow parameters of the steam turbine as well as the environmental and equipment auxiliary parameters that affect heat consumption, to ensure complete capture of the impact of shaft seal leakage, heater heat exchange, and environmental changes on heat consumption.

[0011] Preferably, the threshold for parameter threshold verification in step 1 is set based on the rated operating parameters of the steam turbine. After abnormal data is marked, manual review is triggered. When a sensor failure is confirmed, a backup sensor is started to retransmit data. When instantaneous interference is confirmed, a second completion is performed using the Lagrange interpolation method to ensure the continuity of input data.

[0012] Preferably, in step 1, the state transition matrix of the Kalman filter algorithm is determined based on the dynamic change characteristics of the turbine parameters, the control matrix is ​​set in combination with the characteristics of the load regulation signal, and the Kalman gain is calculated iteratively through the covariance matrix to adapt to the noise removal requirements under different operating conditions.

[0013] Preferably, in step 2, the operating condition identification module divides the load range by calculating the ratio of power generation to the rated power of the turbine (Pe) in real time: a ratio ∈ (25%, 50%) is a low load range, a ratio ∈ (50%, 80%) is a medium load range, and a ratio ∈ (80%, 100%) is a full load range; when the load fluctuation rate is ≥5%Pe / min, the operating condition identification sampling interval is shortened to ensure timely and accurate range judgment.

[0014] Preferably, the historical operating data used to train the random forest correction model in step 2 includes turbine data of different operating years to adapt to the aging characteristics caused by wear of the equipment's flow passage components, and the model training uses cross-validation to ensure generalization ability.

[0015] Preferably, the equipment operating status judgment in step 3 is achieved by comparing the lubricating oil temperature, bearing vibration, and heater terminal difference with the normal operating parameter range of the equipment. If any parameter exceeds the normal range, it is judged as an abnormal state. The abnormality correction coefficient increases with the increase of parameter deviation to ensure that the compensation range matches the degree of parameter abnormality.

[0016] Preferably, the dynamic environmental adaptive correction in step 3 also incorporates the influence of relative humidity. This is achieved by adding a relative humidity correlation term to the calculation of the environmental influence coefficient. When the relative humidity reaches a preset high humidity threshold, the humidity correction weight is increased to adapt to the heat consumption change characteristics under high humidity conditions.

[0017] Preferably, it also includes a heat consumption deviation tracing function: when outputting the final heat consumption value, the deviation degree of each loss item correction coefficient from the standard correction coefficient is output simultaneously. When the deviation degree of a certain loss item exceeds the preset threshold, the turbine equipment component corresponding to the loss item is marked, realizing the association prompt between heat consumption deviation and equipment failure.

[0018] Preferably, for different types of steam turbines, the steam numerical calculation of the mechanism model in step 2 adopts the international water and steam property formulas under the corresponding steam parameters. For ultra-supercritical steam turbines, pressure correction logic is added to the numerical calculation based on the main steam pressure characteristics to adapt to the steam thermodynamic characteristics under ultra-supercritical parameters.

[0019] Advantages of this invention:

[0020] By employing a three-in-one data acquisition system integrating "core + auxiliary + environment," auxiliary data such as shaft seal leakage, low-pressure cylinder exhaust humidity, and heater terminal temperature difference are incorporated, along with environmental data such as ambient temperature and atmospheric pressure. Simultaneously, the DCS sampling frequency is increased to address previous data gaps. Next, a Kalman filter algorithm is used to filter sensor noise, Lagrange interpolation is used to complete missing data, and parameter threshold verification is used to filter out abnormal data. This resolves noise, missing data, and anomaly issues during data acquisition, providing comprehensive and accurate data for heat consumption estimation and fundamentally ensuring the accuracy of the calculation results.

[0021] The heat consumption formula is derived based on thermodynamic principles. Six types of losses are subdivided to construct a quantitative sub-model, which solves the problem of simplifying loss terms in existing models. The model parameters are configured according to the load range through the operating condition identification module to adapt to different operating conditions. Historical data and random forest algorithm are used to optimize the loss correction coefficient. Equipment aging is taken into account to solve the defects of existing models and improve the estimation accuracy.

[0022] By establishing a correlation model between ambient temperature, atmospheric pressure, condenser vacuum, and turbine internal efficiency through regression analysis, and calculating a comprehensive correction coefficient based on equipment operating conditions, dynamic compensation for the impact of environmental factors and equipment status on heat consumption is achieved. This addresses the problems of existing methods that use fixed standard operating parameters and have poor environmental adaptability. Simultaneously, by continuously accumulating comparison results between actual operating data and heat consumption test data, the correlation model parameters are optimized, enabling algorithm self-learning. This ensures that high estimation accuracy is maintained throughout long-term operation, regardless of changes in environmental conditions and equipment status, meeting the heat consumption monitoring needs of different operating scenarios. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 This is a flowchart of the data acquisition and preprocessing process described in this invention;

[0025] Figure 2 This is a flowchart of the preliminary calculation of heat loss and the correction of losses according to the present invention;

[0026] Figure 3 This is a flowchart of the comprehensive correction and verification iteration process described in this invention. Detailed Implementation

[0027] 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.

[0028] like Figure 1 , Figure 2 , Figure 3 As shown, the method for estimating turbine heat consumption based on operating data includes:

[0029] 1. Implementation of Multi-dimensional Data Fusion Acquisition and Preprocessing

[0030] Construction of the acquisition system: Build a three-dimensional acquisition system of "core parameters + auxiliary parameters + environmental parameters". The core parameters are acquired through the distributed control system (DCS) supporting the steam turbine. The sampling frequency of the original DCS system is increased to ensure that the instantaneous changes of parameters under peak load regulation conditions can be captured. The acquired core parameters include main steam pressure, main steam temperature, exhaust pressure, and power generation. The auxiliary parameters are acquired through newly added dedicated sensors. Among them, the shaft seal steam leakage flow is acquired through a shaft seal steam leakage flow sensor, the exhaust humidity of the low-pressure cylinder is acquired through a steam humidity sensor, and the terminal temperature difference of the high-pressure heater is acquired through a temperature difference sensor. The environmental parameters are acquired through an integrated weather station module, specifically including environmental temperature, atmospheric pressure, and relative humidity.

[0031] Execution of data preprocessing: Transmit the various types of original data acquired above to the edge computing node through industrial Ethernet. The edge computing node first runs the Kalman filtering algorithm, presets the state transition matrix based on the dynamic change characteristics of the steam turbine parameters, sets the control matrix in combination with the load regulation signal characteristics, calculates the Kalman gain through covariance matrix iteration, and performs noise rejection processing on the original data to filter out invalid data caused by the self-error of the sensor and electromagnetic interference. For occasional missing data during the acquisition process, the Lagrange interpolation method is used for filling. Based on the valid data points before and after the missing data, the estimated value of the missing data is obtained through interpolation calculation. Finally, parameter threshold verification is performed. The threshold is set based on the rated operating parameters of this subcritical steam turbine. Data outside the threshold range is marked as abnormal data and triggers an artificial review process. If it is confirmed through review that the data abnormality is caused by a sensor failure, a standby sensor is started for data retransmission; if it is confirmed that the abnormality is caused by instantaneous interference, the Lagrange interpolation method is used for secondary data filling, and finally qualified input data is obtained.

[0032] 2. Implementation of Mechanism-Data Fusion-based Refined Heat Consumption Calculation

[0033] Construction of the core formula and loss sub-models: Based on the first law of thermodynamics, the core calculation formula for the heat consumption of the steam turbine is derived as , where D1 is the main steam flow rate, h1 is the main steam enthalpy, D2 is the reheater steam flow rate, h2 is the reheater steam enthalpy, D3 is the exhaust steam flow rate, h3 is the exhaust steam enthalpy, and P is the power generation;

[0034] At the same time, quantitative sub-models for steam flow friction loss, dynamic-static clearance leakage loss, cylinder body heat dissipation loss, valve throttling loss, steam residual velocity loss, and inter-stage steam leakage loss are constructed respectively. Each sub-model is based on the corresponding energy loss mechanism to clarify the correlation between the loss amount and relevant operating parameters.

[0035] Operating Condition Identification and Model Training: The operating condition identification module collects the turbine's power generation in real time, calculates its ratio to the turbine's rated power (Pe), and classifies the load range based on the ratio: a ratio ∈ (25%, 50%) indicates a low load range, a ratio ∈ (50%, 80%) indicates a medium load range, and a ratio ∈ (80%, 100%) indicates a full load range. When the turbine load fluctuation rate is detected to be ≥5% Pe / min, the sampling interval of the operating condition identification module is automatically shortened to ensure the timeliness and accuracy of load range judgment. Historical operating data of this type of subcritical turbine covering various load ranges and different ambient temperatures are collected. The data includes unit operating data from different years of operation to adapt to the aging characteristics caused by wear of equipment flow components. Cross-validation is used to train the historical operating data using a random forest algorithm to optimize the correction coefficients of each loss term, resulting in a random forest correction model.

[0036] Preliminary heat loss calculation: The preprocessed auxiliary parameters are input into the trained random forest correction model, and the model outputs correction coefficients corresponding to each loss term. These correction coefficients are then substituted into the core heat loss calculation formula to correct the results, yielding the preliminary heat loss value. For example, in the low-load range, the correction coefficients for steam flow friction loss and dynamic-static gap leakage loss output by the model differ from those in the medium-load and full-load ranges. This correction process ensures that the preliminary heat loss value reflects the loss characteristics under different loads.

[0037] 3. Implementation of Dynamic Environment Adaptive Correction and Result Output

[0038] Regression analysis was used to analyze the correlation data between historical environmental parameters (ambient temperature, atmospheric pressure, and relative humidity) and condenser vacuum and turbine internal efficiency. A correlation model was established to obtain an environmental impact coefficient, which includes a relative humidity correlation term to reflect the impact of relative humidity on heat consumption. Simultaneously, operating parameters such as turbine lubricating oil temperature, bearing vibration, and heater terminal temperature difference were collected. Each parameter was compared with the normal operating parameter range for this type of turbine. If all parameters were within the normal range, the equipment was considered to be in normal condition, and the anomaly correction coefficient was set to 1. If any parameter exceeded the normal range, it was considered to be in an abnormal condition, and the error was corrected according to the formula… Calculate the anomaly correction coefficient (where ΔX is the deviation between the actual value and the standard value of the parameter, and X0 is the standard value of the parameter). The anomaly correction coefficient increases as the parameter deviation increases.

[0039] The comprehensive correction coefficient is obtained by multiplying the environmental impact coefficient and the anomaly correction coefficient. When the relative humidity of the environment reaches the preset high humidity threshold, the humidity correction weight is automatically increased by 5% to adapt to the heat consumption change characteristics under high humidity environment.

[0040] Final heat consumption value determination and model optimization: Substituting the comprehensive correction coefficient into the initial heat consumption value, the final heat consumption value is calculated. The final heat consumption value is compared and verified with the historical heat consumption data of the turbine. If the deviation is ≤2%, the final heat consumption value is directly output; if the deviation exceeds 2%, the associated model parameters are optimized and the calculation is repeated. Simultaneously, the system continuously accumulates comparison results between actual operating data and heat consumption test data, and periodically optimizes the associated model parameters to achieve self-learning of the algorithm, ensuring the stability of estimation accuracy during long-term operation.

[0041] 4. Implementation of heat consumption deviation traceability function

[0042] While outputting the final heat consumption value, the system simultaneously outputs the deviation of each loss item correction coefficient from the standard correction coefficient. When the deviation of a certain loss item exceeds a preset threshold, the system automatically marks the turbine equipment component corresponding to that loss item. For example, if the deviation of the loss item correction coefficient corresponding to the shaft seal leakage flow exceeds the threshold, the turbine shaft seal is marked, prompting maintenance personnel to check whether the shaft seal has wear, leakage or other faults, thus realizing the correlation between heat consumption deviation and equipment fault.

[0043] 5. Implementation of adaptation for different types of steam turbines

[0044] For supercritical and ultra-supercritical steam turbines, the same estimation process as for subcritical steam turbines is adopted, but adjustments are made in the steam numerical calculation stage. For the steam parameter characteristics of different types of steam turbines, the international water and steam property formulas under the corresponding steam parameters are used to calculate the steam enthalpy values ​​h1, h2, and h3.

[0045] For ultra-supercritical steam turbines, pressure correction logic is added to the numerical calculation based on their main steam pressure characteristics to adapt to the steam thermodynamic characteristics under ultra-supercritical parameters, ensuring that accurate heat consumption estimation results can be obtained for different types of steam turbines.

[0046] Work process:

[0047] 1. Principles of Multi-Dimensional Data Acquisition and Preprocessing

[0048] To address the estimation bias caused by incomplete parameter coverage in traditional methods, this invention constructs a three-dimensional data acquisition system comprising "core parameters + auxiliary parameters + environmental parameters" to comprehensively capture key factors affecting heat consumption. Core parameters focus on the core physical quantities of energy conversion in the steam turbine, including inlet steam parameters, exhaust steam parameters, and unit load. Auxiliary parameters address the completeness of the thermal system, covering extraction parameters at each stage, heater terminal temperature difference, condensate temperature, etc. Environmental parameters address external operating conditions, including ambient temperature, atmospheric pressure, and relative humidity, achieving comprehensive coverage of factors influencing heat consumption.

[0049] The collected raw data needs to be preprocessed by edge computing nodes to ensure the data quality of the input model. A Kalman filter algorithm is used to remove random noise from the sensor data by leveraging its optimal estimation characteristics, and the signal-to-noise ratio is improved by dynamically estimating the system state. To address the issue of missing data, Lagrange interpolation is used to construct an interpolation polynomial based on the effective data sequences before and after the missing data points, accurately repairing missing values ​​to ensure data continuity. A preset threshold verification mechanism is used to filter and remove abnormal data that exceeds reasonable physical ranges or normal fluctuation intervals, ultimately outputting qualified input data with completeness, accuracy, and continuity, laying the foundation for subsequent estimations.

[0050] 2. Mechanism-Data Fusion Type of Refined Modeling Principle

[0051] Based on the first law of thermodynamics (the law of conservation of energy), this study analyzes the energy conversion process of a steam turbine, derives the core calculation formula for heat consumption, clarifies the intrinsic relationship between steam energy input, output, and heat consumption, and defines the calculation logic for steam heat consumption per unit of power generation. Simultaneously, to accurately characterize the energy loss mechanism, six quantitative sub-models for loss terms are constructed, corresponding to inlet throttling losses, intra-stage flow losses, exhaust losses, shaft seal leakage losses, mechanical friction losses, and heat dissipation losses from the thermodynamic system. Through theoretical analysis and fitting of experimental data, quantitative relationships between each loss term and operating parameters are established.

[0052] A working condition identification module is introduced to solve the problem of adapting to all working conditions. The operating range is divided into low, medium, and high load segments based on the unit load signal. For the differences in turbine flow efficiency and loss ratio in different load segments, model parameters (such as loss coefficient and efficiency curve) are configured independently to ensure that the model adapts to the actual operating status of the equipment under each working condition.

[0053] A modified model is constructed by combining the random forest algorithm to achieve deep integration between the mechanistic model and actual operating data. Leveraging the nonlinear fitting capability and high-dimensional data processing advantages of the random forest algorithm, the deviation between the calculated values ​​of the mechanistic model and the actual heat consumption values ​​is used as the training objective. By learning the implicit patterns in a large amount of historical operating data, the correction coefficients of each loss term are optimized, enabling the model to dynamically adapt to performance changes caused by complex factors such as equipment aging and varying operating conditions, thereby improving the model's generalization ability and estimation accuracy.

[0054] 3. Principle of Adaptive Correction for Dynamic Environment and Equipment Status

[0055] To compensate for the impact of environmental changes and equipment condition fluctuations on heat consumption, a two-dimensional correction mechanism is established. On the one hand, a quantitative correlation model is constructed using regression analysis between environmental parameters and condenser vacuum and turbine efficiency. This clarifies the impact patterns of rising ambient temperature leading to decreased condenser vacuum and reduced atmospheric pressure altering condenser back pressure, generating an environmental impact coefficient. On the other hand, real-time monitoring of equipment operating parameters such as bearing temperature, vibration levels, valve opening, and heater terminal temperature difference is used to assess equipment health and performance degradation based on preset condition judgment rules, calculating anomaly correction coefficients. The two types of coefficients are then weighted to obtain a comprehensive correction coefficient, which adjusts the initial heat consumption estimate to ensure the results accurately reflect current operating conditions.

[0056] 4. Algorithm self-learning optimization principle

[0057] A self-learning mechanism is introduced to ensure long-term operational accuracy and stability. The final estimated heat consumption is continuously compared and verified with historical calibration data and benchmark test data of the unit, and the estimation error is calculated. If the error exceeds the preset range, the model parameter optimization process is automatically triggered. Using the latest operating data and error information, the constant terms of the core heat consumption calculation formula, the parameters of the loss sub-model, the weights of the random forest correction model, and the parameters of the correction coefficient associated model are retrained and adjusted. Through continuous feedback optimization, the algorithm can adapt to new operating conditions and equipment status changes, ensuring that high-accuracy heat consumption estimation can be maintained under different environments, loads, and equipment aging stages.

[0058] 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 method for estimating the heat consumption of a steam turbine based on operational data, characterized in that, Includes the following steps: Step 1: Construct a three-dimensional data acquisition system consisting of "core parameters + auxiliary parameters + environmental parameters". Core parameters are acquired through a distributed control system (DCS) with increased sampling frequency, covering main steam pressure, main steam temperature, exhaust pressure, and power generation. Auxiliary parameters are acquired through newly added dedicated sensors, covering shaft seal leakage flow, low-pressure cylinder exhaust humidity, and high-pressure heater terminal temperature difference. Environmental parameters are acquired through a weather station module, covering ambient temperature, atmospheric pressure, and relative humidity. The raw data is transmitted to edge computing nodes, where sensor noise is removed using a Kalman filter algorithm, missing data is completed using Lagrange interpolation, and abnormal data is removed after parameter threshold verification to obtain qualified input data. Step 2: Derive the core calculation formula for turbine heat consumption based on the first law of thermodynamics The model is defined as follows: D1 is the main steam flow rate, h1 is the main steam enthalpy, D2 is the reheat steam flow rate, h2 is the reheat steam enthalpy, D3 is the exhaust steam flow rate, h3 is the exhaust steam enthalpy, and P is the power generation. Quantitative sub-models are constructed to account for steam flow friction loss, dynamic and static clearance leakage loss, cylinder heat dissipation loss, valve throttling loss, steam residual velocity loss, and interstage leakage loss. The turbine load is divided into three intervals—low load, medium load, and full load—through a condition identification module, with model parameters configured independently for each interval. Historical operating data covering each load interval and different ambient temperatures are used, and the random forest algorithm is employed to optimize the correction coefficients for each loss term, resulting in a random forest corrected model. Auxiliary parameters are input into this model, and the model outputs correction coefficients and corrects the core heat consumption calculation formula to obtain the preliminary heat consumption value. Step 3: Establish a correlation model between ambient temperature, atmospheric pressure, condenser vacuum, and turbine internal efficiency through regression analysis to obtain the environmental impact coefficient; collect operating parameters such as turbine lubricating oil temperature, bearing vibration, and heater terminal temperature difference to determine the equipment operating status. Under normal conditions, the abnormal correction coefficient is set to 1; under abnormal conditions, it is adjusted according to... (ΔX is the deviation between the actual value and the standard value of the parameter, and X0 is the standard value of the parameter) Calculate the anomaly correction coefficient, obtain the comprehensive correction coefficient based on the two types of coefficients, substitute it into the preliminary heat consumption value to obtain the final heat consumption value, compare and verify the final heat consumption value with historical data, output if the deviation is ≤2%, and recalculate the model parameters if the deviation exceeds the standard, and accumulate the comparison results of actual operation and heat consumption test data to optimize the associated model parameters to realize the algorithm self-learning.

2. The method for estimating turbine heat consumption based on operating data according to claim 1, characterized in that, In step 1, the three-dimensional acquisition system needs to cover the entire steam flow parameters of the steam turbine as well as the environmental and equipment auxiliary parameters that affect heat consumption, to ensure complete capture of the impact of shaft seal leakage, heater heat exchange, and environmental changes on heat consumption.

3. The method for estimating turbine heat consumption based on operating data according to claim 2, characterized in that, In step 1, the threshold for parameter threshold verification is set based on the rated operating parameters of the steam turbine. Abnormal data is marked and manual review is triggered. When a sensor failure is confirmed, the backup sensor is activated to retransmit data. When instantaneous interference is confirmed, the Lagrange interpolation method is used for secondary completion to ensure the continuity of input data.

4. The method for estimating turbine heat consumption based on operating data according to claim 3, characterized in that, In step 1, the state transition matrix of the Kalman filter algorithm is determined based on the dynamic change characteristics of the turbine parameters, the control matrix is ​​set in combination with the characteristics of the load regulation signal, and the Kalman gain is calculated iteratively through the covariance matrix to adapt to the noise removal requirements under different operating conditions.

5. The method for estimating turbine heat consumption based on operating data according to claim 4, characterized in that, In step 2, the operating condition identification module divides the load range by calculating the ratio of power generation to the rated power of the turbine (Pe) in real time: a ratio of ∈ (25%, 50%) is a low load range, a ratio of ∈ (50%, 80%) is a medium load range, and a ratio of ∈ (80%, 100%) is a full load range; when the load fluctuation rate is ≥5%Pe / min, the operating condition identification sampling interval is shortened to ensure timely and accurate range judgment.

6. The method for estimating turbine heat consumption based on operating data according to claim 5, characterized in that, In step 2, the historical operating data used to train the random forest correction model includes turbine data from different operating years to adapt to the aging characteristics caused by wear of the equipment's flow passage components. Cross-validation is used in the model training to ensure generalization ability.

7. The method for estimating turbine heat consumption based on operating data according to claim 6, characterized in that, In step 3, the equipment operating status is determined by comparing the lubricating oil temperature, bearing vibration, and heater terminal difference with the normal operating parameter range of the equipment. If any parameter exceeds the normal range, it is determined to be an abnormal state. The abnormality correction coefficient increases as the parameter deviation increases, ensuring that the compensation range matches the degree of parameter abnormality.

8. The method for estimating turbine heat consumption based on operating data according to claim 7, characterized in that, In step 3, the dynamic environmental adaptive correction also incorporates the influence of relative humidity. This is achieved by adding a relative humidity correlation term to the calculation of the environmental influence coefficient. When the relative humidity reaches the preset high humidity threshold, the humidity correction weight is increased to adapt to the heat consumption change characteristics under high humidity conditions.

9. The method for estimating turbine heat consumption based on operating data according to claim 8, characterized in that, It also includes a heat consumption deviation tracing function: when outputting the final heat consumption value, it simultaneously outputs the degree of deviation between the correction coefficient of each loss item and the standard correction coefficient. When the deviation of a certain loss item exceeds the preset threshold, it marks the turbine equipment component corresponding to that loss item, realizing the association prompt between heat consumption deviation and equipment failure.

10. The method for estimating turbine heat consumption based on operating data according to claim 9, characterized in that, For different types of steam turbines, the steam numerical calculation of the mechanism model in step 2 adopts the international water and steam property formulas under the corresponding steam parameters. For ultra-supercritical steam turbines, pressure correction logic is added to the numerical calculation based on the main steam pressure characteristics to adapt to the steam thermodynamic characteristics under ultra-supercritical parameters.

Images