A gas turbine state recovery control method based on state evaluation

Abstract

The application relates to the technical field of gas turbines, in particular to a gas turbine state recovery control method based on state evaluation, S1: collecting external environment parameters and monitoring parameters of a gas turbine; S2: establishing a gas turbine simulation twin model by synchronously collecting parameters of a gas turbine physical entity; S3: quantifying the running state of the gas turbine physical entity by a numerical value based on a chromatographic analysis method from three levels of a parameter level, a component level and a device level; S4: running the gas turbine simulation twin model, inputting external environment parameters and related monitoring parameters, setting a target function according to the monitoring parameters, adjusting internal performance parameters of the gas turbine simulation twin model through an optimization algorithm, and achieving synchronization with the gas turbine physical entity; and S5: running state recovery control, selecting input control parameters of the gas turbine, solving optimal input control variables through an optimization algorithm when the running state of the gas turbine is abnormal, and realizing recovery of the running state of the gas turbine.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of gas turbine technology, more particularly to a gas turbine state recovery control method based on state evaluation. BACKGROUND

[0002] Gas turbines are widely used in power generation, aviation and industrial driving fields, and their performance and reliability are crucial to the overall benefits of the system. Gas turbines may degrade in performance or fail due to various factors during long-term operation. Traditional maintenance and control methods mainly rely on periodic inspection and experience-based judgment, which cannot reflect the actual operating state of the gas turbine in real time, resulting in high maintenance costs, slow fault response, and low operating efficiency.

[0003] In order to improve the operating efficiency and reliability of the gas turbine, a recovery control method based on real-time state evaluation is urgently needed. Through real-time evaluation of the state of each component and the overall state of the gas turbine, the external environment, operating conditions and historical data are considered, and the input control parameters are selected and optimized to realize real-time monitoring and intelligent control of the gas turbine, ensuring that it recovers and maintains the best operating state of the gas turbine under the premise of economic efficiency, easy operation and environmental compliance. SUMMARY

[0004] The purpose of the present application is to provide a gas turbine state recovery control method based on state evaluation, which realizes real-time monitoring and intelligent control of the gas turbine through real-time evaluation of the state of each component and the overall state of the gas turbine, and ensures that it recovers and maintains the best operating state of the gas turbine under the premise of economic efficiency, easy operation and environmental compliance.

[0005] The purpose of the present application is achieved by the following technical solutions:

[0006] A gas turbine state recovery control method based on state evaluation, the method comprising the following steps:

[0007] Step S1: Collecting external environmental parameters and monitoring parameters of the gas turbine;

[0008] Step S2: Establishing a gas turbine simulation twin model by synchronizing the parameters of the gas turbine physical entity;

[0009] Step S3: Based on the chromatographic analysis method, quantifying the operating state of the gas turbine physical entity in numerical form from the parameter level, component level and equipment level, and evaluating the operating state of the gas turbine;

[0010] Step S4: Running the gas turbine simulation twin model, inputting the external environmental parameters and related monitoring parameters, setting the target function with the monitoring parameters, adjusting the internal performance parameters of the gas turbine simulation twin model through the optimization algorithm, and achieving synchronization with the gas turbine physical entity.

[0011] Step S5: running state recovery control, selecting input control parameters of the gas turbine, and obtaining optimal input control variables by an optimization algorithm when the running state of the gas turbine is abnormal, so as to realize recovery of the running state of the gas turbine;

[0012] The external environment parameters of the gas turbine in the step S1 include environment temperature, environment pressure and rotating speed, etc., and the monitoring parameters include compressor inlet temperature, compressor inlet pressure, turbine outlet pressure, turbine outlet temperature and fuel flow, etc.

[0013] The method of the gas turbine simulation twin model in the step S2 is a gas turbine simulation model established by using the volume inertia method, and the performance degradation mechanism and the fault factor are introduced to realize performance degradation and fault simulation of the gas turbine.

[0014] The parameter level state evaluation in the step S3 includes three parts, the first part is to establish a monitoring parameter benchmark model, and the normal value of the monitoring parameter under different states is calculated by a regression prediction method; the second part is to calculate the deviation between the normal value and the actual value, and the threshold interval of the monitoring parameter is calculated by a probability density statistical method; the third part is to convert the parameter evaluation result into a certain number between 0 and 1 by a membership function, and the selection of the membership function refers to the trapezoidal and triangular membership functions, and the specific calculation formula is as follows, and the state evaluation result of the parameter level is calculated.

[0015]

[0016] Wherein, a and e are the lower and upper fault thresholds of the parameter respectively; b and e are the lower and upper abnormal thresholds respectively; c is the benchmark value obtained by modeling the benchmark value of the parameter;

[0017] The component level state evaluation in the step S3 is calculated by the radar chart method, that is, the current state of the component is characterized by the area of the polygon, and the state evaluation result of each component is calculated.

[0018] Suppose there is a polygon, and the distance from each vertex to the center point is r1, r2, … r n , and the angle of each vertex is 2Π / n radians; the radar chart is calculated as follows:

[0019]

[0020]

[0021] Wherein, n is the number of vertices, that is, the number of monitoring parameters contained in each component; r nis the evaluation result of each component; S1 is the current component polygon area; Smax is the component polygon area when the theoretical state is best; Acomp is the evaluation result of each component state;

[0022] The step S3 of the equipment level state evaluation integrates the state evaluation results of each component through the Bayesian network to perform overall state evaluation of the equipment. For each node Xi in the Bayesian network, the joint probability density distribution can be expressed as:

[0023]

[0024] Xi represents the i-th node, i.e. all components including the gas turbine and to be evaluated; Pa(Xi) represents the parent node set of Xi;

[0025] The internal performance parameters in the step S4 refer to the internal calculation parameters of the gas turbine simulation twin model, which include the flow correction factor, total pressure recovery coefficient correction factor, and efficiency correction factor, etc.

[0026] The step S4 sets the target function with the monitoring parameters, adjusts the internal performance parameters of the gas turbine simulation twin model through the optimization algorithm, and achieves synchronization with the physical entity of the gas turbine. Specifically, it includes:

[0027] The gas turbine collects monitoring parameters through sensors, including speed, power, exhaust temperature, and compressor outlet pressure; the error c of each target monitoring parameter is calculated i , the simulation calculation value y i of the model, and the actual gas turbine operating parameter value y i is expressed as:

[0028]

[0029] The gas turbine digital twin process is described as an optimization problem of solving a set of component performance parameters to minimize the calculation error of the target monitoring parameter. The objective function is defined as a function of the calculation error of each selected target monitoring parameter:

[0030]

[0031] x={x1,x2,…,x i ,…,x n}

[0032] s.t.x k,min ≤x≤x k,max k=1,2,3,…,n

[0033] In the formula: m is the number of target monitoring parameters; wi a weighting coefficient for each target monitoring parameter, n is the number of performance parameters; x i is a performance parameter; x k,max and x k,min are respectively the upper limit value and the lower limit value of the corresponding performance parameter correction factor;

[0034] The optimization algorithm specifically comprises: a particle swarm optimization algorithm based on particle forbidden zones and Levy flight is adopted, a traditional PSO update and a Levy flight update are balanced through a probability mechanism, particles are redistributed by setting particle forbidden zones, particles located in the forbidden zone at the beginning are redistributed to the position of the previous particle, and particles entering the forbidden zone in the iteration process retain the position of the last iteration;

[0035] The input control parameter of the gas turbine is selected in the step S4, and specifically comprises: the state recovery control of the gas turbine needs to comprehensively consider the state evaluation results of each component, environmental conditions, operating conditions, historical data, economy, operability and environmental protection; by selecting appropriate input control parameters, the gas turbine is ensured to recover normal operation under the premise of economic efficiency, simple operation and environmental compliance;

[0036] In the step S5, when the operating state of the gas turbine is abnormal, the optimal input control amount is obtained through the optimization algorithm to realize the recovery of the operating state of the gas turbine;

[0037] Specifically comprising: by adjusting the input control amount, the operating state evaluation result of the gas turbine is maximized, and the specific calculation is as follows:

[0038] Obj = min{1 - Assessment[gasturbin(ΔIN1, ΔIN2,..., ΔIN n )]

[0039] In the formula, gasturbine is a twin model of the gas turbine, the input is the input control amount of the gas turbine, the output is the real-time monitoring parameter of the gas turbine, Assessment is an operating state evaluation model, the input is the monitoring parameter of the gas turbine, and the output is the current state evaluation result of the gas turbine; ΔIN n is the input control amount.

[0040] The beneficial effects of the present application are:

[0041] The simulation twin model of the gas turbine is established by the volumetric inertia method, the performance degradation mechanism and the fault factors are introduced, the simulation of the performance degradation and the fault of the gas turbine is realized, the internal performance parameters of the simulation twin model of the gas turbine are adjusted through the optimization algorithm, the simulation twin model is synchronized with the physical entity of the gas turbine, the virtual and real combination is realized, the practicability and the reliability of the simulation model are enhanced, the performance parameters are directly reflected on the running state of the gas turbine through the twin mechanism; the state assessment is carried out from the parameter level, the component level and the equipment level, the state assessment is carried out layer by layer, and each aspect of the gas turbine is comprehensively covered, so that the comprehensiveness and the reliability of the assessment result are ensured; the control recovery parameters of the gas turbine are determined, when the running state of the gas turbine is abnormal, the optimal input control quantity is solved through the optimization algorithm by taking the input control quantity as the independent variable and the state assessment result as the dependent variable, and the target optimization of the running state of the gas turbine is carried out. BRIEF DESCRIPTION OF DRAWINGS

[0042] The application will be further described in detail below in combination with the drawings and specific implementation methods.

[0043] Fig. 1 is a state recovery control method for a gas turbine based on state assessment;

[0044] Fig. 2 is a state recovery control method for a gas turbine based on state assessment;

[0045] Fig. 3 is a state recovery result diagram of the gas turbine. DETAILED DESCRIPTION

[0046] The application will be further described in detail below in combination with the drawings.

[0047] As shown in the drawings, Figs. 1-3 in order to realize the technical effect of "realizing the real-time monitoring and intelligent control of the gas turbine through the real-time assessment of the states of each component and the whole of the gas turbine, and ensuring the recovery and maintenance of the optimal running state of the gas turbine under the premise of economic efficiency, simple operation and environmental compliance", the steps and functions of the state recovery control method for the gas turbine based on state assessment are described in detail below.

[0048] As shown in the drawings, Fig. 1 a state recovery control method for a gas turbine based on state assessment, the method comprising the following steps:

[0049] Step S1: data acquisition, acquiring the external environmental parameters and the monitoring parameters of the gas turbine;

[0050] In the specific implementation, the external environmental parameters and the monitoring parameters of the gas turbine are dynamically acquired through the sensor installed on the gas turbine and the acquisition unit of the gas turbine control system.

[0051] Taking a certain type of three-shaft gas turbine as an example, the collected external environment parameters and monitoring parameters mainly include: ambient temperature, ambient pressure, low-pressure compressor inlet temperature, low-pressure compressor inlet pressure, low-pressure compressor outlet temperature, low-pressure compressor outlet pressure, high-pressure compressor outlet temperature, high-pressure compressor outlet pressure, low-pressure turbine outlet temperature, low-pressure turbine outlet temperature, power turbine outlet temperature, power turbine outlet pressure, low-pressure turbine speed, high-pressure turbine speed, power turbine speed, fuel flow, throttle opening, etc.

[0052] Step S2: Establishing a simulation model, a gas turbine simulation twin model is established through the parameters of the physical entity of the gas turbine;

[0053] In specific embodiments, the volume inertia method is used to establish a gas turbine simulation model. Taking the gas path components of a certain type of three-shaft gas turbine as an example, the simulation model includes a low-pressure compressor, a high-pressure compressor, a combustor, a high-pressure turbine, a low-pressure turbine, a power turbine, etc.

[0054] Step S3: Running state evaluation, based on the chromatographic analysis method, the running state of the physical entity of the gas turbine is quantified in numerical value from the parameter level, component level and equipment level, and the running state of the gas turbine is evaluated;

[0055] In specific embodiments, a monitoring parameter benchmark model is established by a long short-term memory network. The input parameters of the benchmark model are ambient temperature, ambient pressure and calculated torque. The normal values of the monitoring parameters under different states are calculated by a regression prediction method based on LSTM.

[0056] In specific examples, the parameters are normalized by the maximum and minimum value method. i,j is the original data, is the normalized value of the i-th parameter j. and is the maximum and minimum value of parameter j. Data preprocessing normalizes the data to [0, 1], reducing the influence of working conditions and environment.

[0057]

[0058] In specific embodiments, historical data of the input window size is used for prediction. The window length is set to n, so the two-dimensional input dimension of the three input parameters constructed is (3, n), and the corresponding output dimension is (1, p). That is, n steps of historical data are used to predict the trend p steps later. Wherein, n = 3, p = 1, input into the LSTM model, train the LSTM benchmark prediction model, and realize the prediction of the future 1 step benchmark value.

[0059] In specific embodiments, the baseline model is composed of 2 layers of LSTM layers and 1 layer of fully connected layers. In the LSTM hidden layer, the low-pressure turbine exhaust temperature time series features are extracted, and then the regression from features to prediction values is realized through the fully connected layer. The tanh activation function, L2 regularization (coefficient 0.001), Adam optimizer (learning rate 0.01), mse loss function, and 300 training iterations are used.

[0060] In specific embodiments, the threshold interval of the monitoring parameter is calculated by the method of kernel density estimation, and the selection of the threshold interval can be set by the confidence interval, wherein the upper and lower limit abnormal thresholds are 95% confidence intervals, and the upper and lower limit fault thresholds are 99% confidence intervals.

[0061] In specific embodiments, the result of parameter evaluation is converted to a certain number between 0 and 1 by using the membership function, and the selection of the membership function refers to the trapezoidal and triangular membership functions. The specific calculation formula is as follows, and the state evaluation result of the parameter level is calculated.

[0062]

[0063] Wherein, a and e values are the lower limit fault threshold and the upper limit fault threshold of the parameter; b and e values are the lower limit abnormal threshold and the upper limit abnormal threshold; c is the reference value obtained by modeling the reference value of the parameter.

[0064] In specific embodiments, the gas turbine contains 4 components, which are represented by 4 vertices in the radar chart. The distance from each vertex to the center point is r1, r2, r3, and r4, respectively, and the angle of each vertex is 90 degrees. The radar chart is calculated as follows.

[0065]

[0066] In specific embodiments, the gas turbine is divided into 4 components, namely the gas path, the fuel system, the oil system, and the cooling system. The Bayesian network contains 5 nodes, and the joint probability density distribution can be represented as:

[0067] P(T,G,O,F,C)=P(T|G,O,F,C)·P(G)·P(O)·P(F)·P(C)

[0068] Wherein, T represents the gas turbine, and G, O, F, and C represent the gas path, the fuel system, the oil system, and the cooling system, respectively.

[0069] Step S4: Run the gas turbine simulation twin model, input the external environmental parameters and related monitoring parameters, set the target function with the monitoring parameters, and adjust the internal performance parameters of the gas turbine simulation twin model through the optimization algorithm to achieve synchronization with the physical entity of the gas turbine.

[0070] In specific embodiments, taking the gas path components as an example, the internal performance parameters include: low-pressure compressor flow correction factor, low-pressure compressor efficiency correction factor, high-pressure compressor flow correction factor, high-pressure compressor efficiency correction factor, combustor total pressure recovery coefficient correction factor, high-pressure turbine flow correction factor, high-pressure turbine efficiency correction factor, low-pressure turbine flow correction factor, low-pressure turbine efficiency correction factor, power turbine flow correction factor, and power turbine efficiency correction factor.

[0071] In specific embodiments, the optimization algorithm has a particle swarm size N of 30, an iteration number iter of 500 times, a search space dimension dim of 10, an inertia weight w of 0.7, a cognitive coefficient c1 of 1.2, a social coefficient c2 of 1.5, and a Levy flight step length RL of 0.15*levy(N, dim, 1.5), wherein 1.5 is the beta parameter of the Levy distribution.

[0072] Step S5: running state recovery control, selecting the input control parameters of the gas turbine, and when the running state of the gas turbine is abnormal, obtaining the optimal input control amount through the optimization algorithm to realize the recovery of the running state of the gas turbine;

[0073] Specifically, the input control amount is adjusted to maximize the evaluation result of the running state of the gas turbine, and the specific calculation is as follows:

[0074] Obj = min{1 - Assessment[gasturbine(ΔIN1, ΔIN2, …, ΔIN n )]}

[0075] In the formula, gasturbine is a twin model of the gas turbine, the input is the input control amount of the gas turbine, the output is the real-time monitoring parameter of the gas turbine, Assessment is a running state evaluation model, the input is the monitoring parameter of the gas turbine, and the output is the current state evaluation result of the gas turbine; ΔIN n is the input control amount;

[0076] In specific embodiments, the control speed and the intake flow are selected as the input control amount.

[0077] In specific embodiments, since the speed range of the power turbine of the gas turbine is between 3000 r / min and 3600 r / min, the upper and lower limits of the change amount of the control speed parameter of the gas turbine are determined as [-600, 600] with the unit of r / min. Due to the adjustable guide vane restriction, the change amount of the intake flow should be controlled within 10%, and the intake flow is about 80 kg / s under the 0.95 working condition. The upper and lower limits of the change amount of the intake flow parameter are determined as [-6, 6] with the unit of kg / s.

[0078] In specific embodiments, the recovery states reached under different faults are also different. Among them, fault 5 has the best recovery effect, from the initial state of health 0.5846 to 0.9795. Fault 15 has the worst recovery effect, from the initial state of health 0.4602 to 0.4921, which can only slightly improve the performance of the gas turbine. Overall, the recovery results under different states are improved.

[0079] In specific embodiments, the optimized input control variables are as shown in the following table:

[0080] Table 5.3 Input control variables under fault state

[0081]

[0082] It should be noted that the steps shown in the above process or the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that here.

[0083] The above embodiments only express several embodiments of the present application, and the description is more specific and detailed, but it should not be understood as a limitation on the scope of the patent. It should be noted that for ordinary skilled in the art, without departing from the concept of the present application, a number of modifications and improvements can be made, which are within the scope of the present application. Therefore, the scope of protection of the patent of the present application should be subject to the appended claims.

Claims

1. A gas turbine state recovery control method based on state assessment, characterized by: The method comprises the following steps: Step S1: collecting gas turbine external environment parameters and monitoring parameters; Step S2: Synchronously establishing a gas turbine simulation twin model through parameters of the gas turbine physical entity; Step S3: Based on the tomographic analysis method, the operating status of the gas turbine physical entity is numerically quantified from three levels: parameter level, component level, and equipment level, to evaluate the operating status of the gas turbine; Step S4: Run the gas turbine simulation twin model, input external environmental parameters and related monitoring parameters, set the objective function based on the monitoring parameters, and adjust the internal performance parameters of the gas turbine simulation twin model through the optimization algorithm to achieve synchronization with the physical entity of the gas turbine; In step S4, the objective function is set based on the monitoring parameters, and the internal performance parameters of the gas turbine simulation twin model are adjusted through the optimization algorithm to achieve synchronization with the physical entity of the gas turbine, which specifically includes: The gas turbine collects monitoring parameters through sensors, including speed, power, exhaust temperature and compressor outlet pressure; the calculation error c of each target monitoring parameter is i , using the model's simulation calculation value y i ′ and the actual gas turbine operating parameter value y i Expressed as: The gas turbine digital twin process is described as an optimization problem to solve a set of component performance parameters so that the calculation error of the target monitoring parameters is minimized. The objective function is defined as a function of the calculation error of each selected target monitoring parameter: x={x1,x2,…,x i ,…,x n } s.t.x k,min ≤x≤x k,max k=1,2,3,…,n Where: m is the number of target monitoring parameters; w i is the weighting coefficient of each target monitoring parameter, n is the number of performance parameters; x i is the performance parameter; x k,max and x k,min are the upper and lower limits of the corresponding performance parameter correction factors respectively; Step S5: operating state recovery control, selecting the input control parameters of the gas turbine, and obtaining the optimal input control amount through an optimization algorithm when an abnormal operating state of the gas turbine occurs, so as to achieve recovery of the operating state of the gas turbine; In step S5, when the operating state of the gas turbine is abnormal, an optimal input control variable is obtained by solving the optimization algorithm to restore the operating state of the gas turbine; Specifically, the input control amount is adjusted to maximize the gas turbine operating status evaluation result. The specific calculation is as follows: Obj=min{1-Assessment[gasturbin(ΔIN1,ΔIN2,...,ΔIN n )]}; Where, gasturbin is the gas turbine twin model, the input is the gas turbine input control variable, and the output is the gas turbine real-time monitoring parameter; Assessment is the operation status assessment model, the input is the gas turbine monitoring parameter, and the output is the current gas turbine status assessment result; ΔIN n is the input control quantity.

2. A gas turbine state recovery control method based on state assessment according to claim 1, characterized in that: The external environmental parameters of the gas turbine in step S1 include: ambient temperature, ambient pressure and rotation speed; the monitoring parameters include: compressor inlet temperature, compressor inlet pressure, turbine outlet pressure, turbine outlet temperature and fuel flow.

3. The gas turbine state recovery control method based on state assessment according to claim 1, characterized in that: The method for simulating the twin model of the gas turbine in step S2 is to establish a gas turbine simulation model using the volume inertia method, and to achieve gas turbine performance degradation and failure simulation by introducing performance degradation mechanisms and failure factors.

4. The gas turbine state recovery control method based on state assessment according to claim 1, characterized in that: The parameter-level status assessment in step S3 consists of three parts: the first part establishes a monitoring parameter benchmark model and calculates the normal values ​​of the monitoring parameters under different states by using a regression prediction method; The second part calculates the deviation between the normal value and the actual value, and obtains the threshold interval of the monitoring parameter by probability density statistics; The third part calculates the parameter-level state assessment results through the membership function.

5. The gas turbine state recovery control method based on state assessment according to claim 1, characterized in that: In step S3, the component-level status assessment is performed by using a radar chart method, i.e., polygonal area represents the current status of the component, and the status assessment result of each component is calculated; Suppose there is a polygon, and the distance from each vertex to the center point is r1, r2, ... r n , and the angle of each vertex is 2Π / n radians; the radar chart is calculated as follows: Where n is the number of vertices, that is, the number of monitoring parameters contained in each component; r n is the distance from each vertex to the center point, that is, the evaluation result of each monitoring parameter; S1 is the polygonal area of ​​the current component; Smax is the polygonal area of ​​the component when the theoretical state is the best; Acomp is the evaluation result of each component state.

6. The gas turbine state recovery control method based on state assessment according to claim 1, characterized in that: In the device-level status assessment in step S3, the status assessment results of each component are integrated through the Bayesian network to perform an overall status assessment of the device.

7. The gas turbine state recovery control method based on state assessment according to claim 1, characterized in that: The internal performance parameters in step S4 refer to the internal calculation parameters of the gas turbine simulation twin model, and the internal calculation parameters of the gas turbine simulation twin model include a flow correction factor, a total pressure recovery coefficient correction factor, and an efficiency correction factor.

8. The gas turbine state recovery control method based on state assessment according to claim 1, characterized in that: The input control parameters of the gas turbine are selected in step S4, specifically including: the state recovery control of the gas turbine needs to comprehensively consider the state assessment results of each component, environmental conditions, operating conditions, historical data, economy, operability and environmental protection.

Images