Method for predicting health state of exhaust temperature of gas turbine

Abstract

The invention discloses a method for predicting the health state of the exhaust temperature of a gas turbine. The method comprises the steps that 1, a fault-free exhaust temperature data set Tt and a fault-free working condition data set Wt of the gas turbine are collected under different stable operation working conditions; step 2, performing data processing on the Wt; 3, selecting working condition characteristics same as Wt, and collecting a group of predicted working condition vectors Wp and predicted exhaust temperature vectors Tp in real time; step 4, performing data processing on the Wp, and then calculating a correlation coefficient between the Wp and the Wt after data processing to obtain a working condition coefficient vector C; 5, calculating the distance between the Tt and the Tp to serve as an exhaust temperature distance vector D, and converting the exhaust temperature distance vector D into an exhaust temperature similarity vector S; and step 6, according to the working condition coefficient vector C and the exhaust temperature similarity vector S, calculating to obtain the health degree h of the data sample to be predicted so as to quantify the health state of the exhaust temperature of the gas turbine.

Description

Technical Field

[0001] The invention relates to a method for monitoring a gas turbine, and in particular to a method for predicting the healthy state of exhaust temperature of a gas turbine. Background Art

[0002] The temperature before the turbine of a heavy-duty gas turbine is very high, and the high temperature poses a great challenge to the reliability of the hot end components of a heavy-duty gas turbine. The current industrial sensor technology is difficult to directly measure the temperature of high-temperature components reliably for a long time. In practical applications, the turbine exhaust temperature is generally measured by evenly arranging several thermocouples in the turbine exhaust passage to indirectly monitor the status of the high-temperature hot end components. When the combustion is unstable, the combustion chamber or the transition section is broken, it may cause abnormal inlet and exhaust temperatures of the turbine and imbalance of the temperature flow field. However, due to noise interference or the limitations of the sensor itself, the temperatures measured by each thermocouple cannot be absolutely the same. In addition, when the operating conditions such as the load of the gas turbine change, the exhaust temperature will also change. Therefore, how to use the measured turbine exhaust temperature data to construct appropriate indicators or analysis methods to accurately reflect the health status of high-temperature hot channel components is a critical research topic at present.

[0003] Foreign OEM companies such as GE and Siemens have set up a series of protection logic mechanisms for exhaust temperature. Siemens installed 24 K-graded three-wire thermocouples in the exhaust diffusion section of heavy-duty gas turbines, and selected 6 representative measurement points for protection configuration, protection action and alarm setting: when the temperature of any three points is higher than 659 degrees Celsius (exhaust temperature correction value), the unit will trigger the protection action and trip; if the sensor signals of any three points or more are broken, the unit will enter the shutdown procedure.

[0004] GE has arranged 31 thermocouples along the circumferential direction in the rear channel of the exhaust diffuser section of the gas turbine. When the unit is running under stable conditions, the exhaust temperature field cannot be completely uniform, and the readings of each thermocouple are different. It is necessary to determine the temperature difference allowed to be measured by each thermocouple when the unit is operating normally, which is called the allowable exhaust dispersion S. allow When the engine is running, the exhaust temperature values ​​measured by 31 exhaust thermocouples are sent to the control system. The temperature value sequence is obtained from high to low, and the difference between the highest exhaust temperature and the lowest exhaust temperature is calculated as S1, the difference between the highest exhaust temperature and the second lowest exhaust temperature is S2, and the difference between the highest exhaust temperature and the third lowest exhaust temperature is S3. The values ​​of these three actual dispersions are compared with the values ​​of the allowable dispersion. If S1 / S allow ≥1, and S2 / S allow ≥0.8, and the lowest and second lowest exhaust temperatures are adjacent to each other, the unit will be shut down; if S1 / S allow ≥5, and S2 / S allow≥0.8 and the second lowest and third lowest exhaust temperatures are adjacent, the unit will be shut down; if S3 / S allow ≥0.75, the unit will be shut down.

[0005] The monitoring logic of gas turbine exhaust temperature adopted by GE and Siemens is based on the operating data and operating status of each gas turbine collected by OEM for a long time, and the thresholds and judgment rules obtained by refining and summarizing the rules and experience. By using some simple logical judgment rules and setting relatively fixed temperature thresholds or dispersion thresholds, the combustion monitoring and protection of gas turbines can be carried out. However, this empirical threshold determination rule is not suitable for newly developed heavy-duty gas turbines on the one hand, and on the other hand, it cannot be dynamically and adaptively adjusted according to the actual operating conditions of the gas turbine power plant, weather and climate changes, actual start-up and shutdown and maintenance status of the gas turbine, and cannot achieve the role of intelligent monitoring or early warning.

[0006] In addition, patent CN103195583B proposes a method for protecting gas turbine combustion by exhaust gas dispersion monitoring. The method mainly uses the multi-dimensional space cosine theorem to obtain the turbine exhaust temperature dispersion index, and on this basis, a threshold is established for monitoring and alarming. However, through the study of a large amount of actual gas turbine operation data, it is found that when the operating conditions of the gas turbine change, the weather and climate change, or the combustion chamber is overhauled, the exhaust temperature data measured by dozens of temperature measuring thermocouples of the gas turbine turbine may not change linearly and synchronously. At this time, the combustion in the combustion chamber is still stable and trouble-free, and the calculated exhaust temperature dispersion index may be lower than the set threshold, thereby triggering an error alarm.

[0007] In view of the above technical problems, the present invention is specially introduced. Summary of the invention

[0008] The main purpose of the present invention is to provide a method for predicting the healthy state of the exhaust temperature of a gas turbine, which is used to propose a prediction method that has the ability to adaptively adjust to different combustion chamber designs, different stable operating conditions, and different climate and environmental changes.

[0009] In order to achieve the above object, the present invention proposes a method for predicting the health status of gas turbine exhaust temperature, comprising the following steps: Step 1, collecting data sets of gas turbine trouble-free operation under different stable operating conditions, including a trouble-free exhaust temperature data set T t and the fault-free condition data set W t ; Step 2, for the fault-free working condition data set W t Perform data processing; Step 3, select the data set W with the fault-free working condition t For the same working condition characteristics, a set of data samples to be predicted are collected in real time, including the predicted working condition vector W pand the predicted exhaust temperature vector T p ; Step 4, predict the operating condition vector W p Perform data processing and then calculate the fault-free working condition data set W after data processing t The correlation coefficient between them is obtained to obtain the operating coefficient vector C; Step 5, calculate the fault-free exhaust temperature data set T t and the predicted exhaust temperature vector T p The distance between them is used as the exhaust temperature distance vector D, and is converted into the exhaust temperature similarity vector S; Step six, according to the operating condition coefficient vector C and the exhaust temperature similarity vector S, the health h of the data sample to be predicted is calculated to quantify the health state of the gas turbine exhaust temperature.

[0010] Furthermore, in step 5, the predicted exhaust temperature vector T p The elements in are twisted and sorted in clockwise or counterclockwise order to obtain a set of twisted exhaust temperature vectors, and then twisted multiple times in order to obtain multiple sets of exhaust temperature vectors T p_i .

[0011] Further, in step 5, the fault-free exhaust temperature data set T t Each fault-free exhaust temperature vector T t_j Respectively with multiple groups of exhaust temperature vectors T p_i The minimum value of the distance is calculated to obtain the exhaust temperature distance vector D.

[0012] Furthermore, in step five, the exhaust temperature distance vector D is converted into an exhaust temperature similarity vector S by constructing a kernel function, an exponential function or a logarithmic function.

[0013] Furthermore, in step six, each element in the operating condition coefficient vector C is multiplied by the corresponding element in the exhaust temperature similarity vector S, and the maximum value is taken as the health h of the data sample.

[0014] Furthermore, the element 0≤s in the exhaust temperature similarity vector S j ≤1, the element 0≤c in the operating coefficient vector C j ≤1.

[0015] Furthermore, the exhaust temperature vector T is predicted p There are k elements in it, which are the exhaust temperature values ​​measured by k thermocouples distributed along the circumference. p Perform k-1 temperature twists to obtain k sets of exhaust temperature vectors T p_i .

[0016] Furthermore, in step 2, the fault-free working condition data set W is analyzed according to the characteristics of each working condition. tPerform data normalization to obtain the normalized working condition data set W′ t , and the normalized model function γ.

[0017] Furthermore, in step 4, the normalized model function γ is used to predict the operating condition vector W p Perform data normalization to obtain the normalized working condition vector W′ p , and calculate W′ p and W′ t The correlation coefficient of the operating condition coefficient vector C is obtained.

[0018] Furthermore, in step 2, the normalization algorithm uses the minimum-maximum normalization or the decimal calibration normalization algorithm, so that the normalized working condition data set W′ t The values ​​of each operating condition feature in are distributed in the range of [0, 1].

[0019] Furthermore, in step 4, the normalized working condition vector W′ is calculated respectively p With the normalized working condition data set W′ t The absolute value of the Pearson correlation coefficient of each vector in is used to obtain the operating condition coefficient vector C.

[0020] Furthermore, the operating condition characteristics may be selected from: several or all of the following: rotation speed, load, IGV opening, compressor intake pressure, fuel temperature, fuel Weber index, ambient temperature, ambient atmospheric pressure, and ambient humidity.

[0021] By applying the technical solution of the present invention, at least the following beneficial effects are achieved:

[0022] 1. After selecting a section of fault-free data with different operating conditions, the prediction method of the present invention selects key operating condition features to form a fault-free operating condition data set and a fault-free exhaust temperature data set, fully considering the impact of the operating conditions on the exhaust temperature. By collecting more fault-free data sets that cover different operating conditions as much as possible, the method has adaptive adjustment capabilities in predicting the exhaust temperature health under different stable operating conditions and weather conditions.

[0023] 2. The prediction method of the present invention normalizes the fault-free operating condition data set and normalizes the predicted operating condition vector using the recorded normalization model, thereby facilitating the calculation of the correlation between the predicted operating condition vector and the fault-free operating condition data set to obtain the operating condition coefficient vector for health correction.

[0024] 3. The prediction method of the present invention simulates the torsion of the temperature field, twists and sorts the elements of the exhaust temperature vector in sequence, obtains a new set of exhaust temperature vectors, and uses this method to obtain multiple sets of exhaust temperature vectors, and calculates the minimum value of the distance between them and each sample of the fault-free exhaust temperature data set. The prediction method of the present invention fully considers the torsion angle between the exhaust temperature distribution and the actual corresponding combustion chamber, so that the prediction method has adaptive ability to predict the exhaust temperature health of different combustion chamber designs.

[0025] 4. The prediction method of the present invention converts the exhaust temperature distance vector into an exhaust temperature similarity vector through a kernel function, and designs a calculation formula for the health of data samples, thereby scientifically quantifying the health status of the gas turbine exhaust temperature based on the similarity calculation method. The method has been proven to be effective in practice and is easy to verify the design of algorithms and parameters to improve the accuracy and sensitivity of the algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The accompanying drawings constituting a part of the present invention are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the accompanying drawings:

[0027] Figure 1 A flow chart of a method for predicting the healthy state of exhaust temperature of a gas turbine according to the present invention is shown;

[0028] Figure 2 A schematic diagram showing the exhaust temperature field of the present invention being twisted once is shown. DETAILED DESCRIPTION

[0029] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments can be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0030] The present invention is further described in detail below in conjunction with specific embodiments, and these embodiments cannot be understood as limiting the scope of protection claimed by the present invention. The term "including" when used indicates the existence of a feature, but does not exclude the existence or addition of one or more other features; the orientation or position relationship indicated by the terms "lateral", "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inside", "outside", etc. is based on the orientation or position relationship shown in the drawings, which is only for the convenience of description, and does not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention; in addition, the terms "first" and "second" are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance.

[0031] In the description, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be directly connected, or indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances. In addition, in the description of the present invention, unless otherwise specified, "plurality" means two or more.

[0032] Example:

[0033] The present invention proposes a method for predicting the healthy state of the exhaust temperature of a gas turbine. Figure 1 As shown, the specific steps include:

[0034] Step 1: Collect data sets of gas turbine trouble-free operation under different stable operating conditions, including trouble-free exhaust temperature data sets T t and the fault-free condition data set W t These data must be collected when the gas turbine is in good condition and for a long period of time, covering different stable operating conditions and weather changes.

[0035]

[0036] Where m is the number of samples of the fault-free exhaust temperature vector group collected, which is related to the collection time and sampling frequency; k represents the exhaust temperature value measured by k thermocouples distributed circumferentially in the combustion chamber; T t_1 ,T t_2 ,……T t_m It is the abbreviation of m fault-free exhaust temperature vector groups.

[0037]

[0038] Among them, m is the number of samples of the fault-free working condition vector group collected, which is related to the collection time and sampling frequency; w1,w2,w3,…w n is n operating condition characteristics, and the operating condition characteristics can be selected from: speed, load, IGV opening, compressor inlet pressure, fuel temperature, fuel Weber index, ambient temperature, ambient atmospheric pressure, ambient humidity, etc., or some or all of them; W t_1 ,W t_2 ,……W t_m It is the abbreviation of m fault-free operating condition vector groups.

[0039] The prediction method of the present invention selects a section of fault-free data with different operating conditions, selects key operating condition features to form a fault-free operating condition data set and a fault-free exhaust temperature data set, and fully considers the impact of the operating conditions on the exhaust temperature. By collecting more fault-free data sets that cover different operating conditions as much as possible, the method has adaptive adjustment capabilities in predicting the health of exhaust temperature under different stable operating conditions and weather conditions.

[0040] Step 2: For the fault-free working condition data set W t According to the characteristics of each working condition, that is, the data is normalized according to each column to obtain the normalized working condition data set W′ t , and the normalized model function γ.

[0041] Preferably, the normalization algorithm adopts a minimum-maximum normalization or a decimal calibration normalization algorithm, so that the normalized working condition data set W′ t The values ​​of each operating condition feature in are distributed in the range of [0, 1].

[0042]

[0043] Step 3: Select the data set W with no fault condition t For the same working condition characteristics, a set of data samples to be predicted are collected in real time, including the predicted working condition vector W p and the predicted exhaust temperature vector T p .

[0044] T p =[t 1p t 2p …t kp ]

[0045] Among them, t 1p ,t 2p …t kp is the newly collected t The same k thermocouple sensors measure a set of exhaust temperature data.

[0046] W p =[w 1p w 2p …w n p]

[0047] Among them, W 1p ,W 2p ,…,W np is the newly collected t A group of operating condition data consisting of n identical operating condition characteristics.

[0048] Step 4: Use the normalized model function γ to predict the operating condition vector W pAfter processing, the normalized working condition vector W′ is obtained p , and calculate W′ p and W′ t The correlation coefficient of the operating condition coefficient vector C is obtained.

[0049] W′ p =γ(W p )=[w′ 1p w′ 2p …w′ np ]

[0050] Furthermore, it is preferred to calculate the normalized working condition vector W′ p With the normalized working condition data set W′ t The absolute value of the Pearson correlation coefficient of each vector in is obtained to obtain the operating coefficient vector C. Other methods for calculating the correlation coefficient can also be used.

[0051]

[0052] Among them, c1, c2, …, c m ∈[0,1].

[0053] By normalizing the fault-free operating condition data set and normalizing the predicted operating condition vector using the recorded normalization model, it is convenient to calculate the correlation between the predicted operating condition vector and the fault-free operating condition data set to obtain the operating condition coefficient vector for health correction.

[0054] Step 5: Calculate the fault-free exhaust temperature data set T t and the predicted exhaust temperature vector T p The distance between them is obtained to obtain the exhaust temperature distance vector D, which is then converted into the exhaust temperature similarity vector S.

[0055] According to relevant research, the distribution of the gas turbine exhaust temperature field is closely related to the combustion conditions of each combustion chamber. Each combustion chamber will relatively fixedly correspond to several measurement points of the gas turbine exhaust temperature. When the gas turbine load changes, the measured exhaust temperature distribution may have a certain twist angle with the actual corresponding combustion chamber. Therefore, when calculating the health, the temperature twist should be taken into account. Rotate k exhaust temperature distributions clockwise k times, and each rotation angle is 360° / k. As shown in the attached figure Figure 2 As shown, taking k = 12 thermocouples as an example, Figure 2 The middle left picture shows the exhaust temperature distribution before twisting, and the right picture shows the exhaust temperature distribution after one clockwise twist.

[0056] The predicted exhaust temperature vector T p Set to T p_1 , T p_1The elements in are twisted and sorted in clockwise or counterclockwise order. Taking clockwise as an example, the following set of twisted exhaust temperature vectors T are obtained. p_2 .

[0057] T p_1 =[t 1p t 2p … t (k-1)p t kp ]

[0058] T p_2 =[t 2p t 3p … t kp t 1p ]

[0059] Continue to twist clockwise multiple times to obtain multiple sets of exhaust temperature vectors T p_i as follows:

[0060]

[0061] Preferably, T p_1 Perform k-1 temperature twists to obtain k sets of exhaust temperature vectors T p_i .

[0062] Unfaulted exhaust temperature data set T t Each fault-free exhaust temperature vector T t_j Respectively with the above multiple groups of exhaust temperature vectors T p_i Calculate the minimum value of the distance to obtain the exhaust temperature distance vector D, which is used as the predicted temperature vector T p Compared with the fault-free exhaust temperature data set T t The distance between.

[0063]

[0064] Among them, Distance is the distance calculation function, and the methods that can be used include common distance algorithms such as cosine distance and Euclidean distance. Take Euclidean distance as an example to solve the distance of a set of twisted temperature data vector distribution:

[0065]

[0066] The prediction method of the present invention fully considers the torsion angle between the exhaust temperature distribution and the actual corresponding combustion chamber, so that the prediction method has adaptive ability to predict the exhaust temperature health of different combustion chamber designs.

[0067] Next, the exhaust temperature distance vector D is converted into an exhaust temperature similarity vector S by constructing a kernel function, an exponential function or a logarithmic function.

[0068]

[0069] Preferably, is a method to convert the distance d j Convert to s in the value range [0,1] j Gaussian kernel function.

[0070] Step six, based on the operating coefficient vector C and the exhaust temperature similarity vector S, calculate the health h of the data sample to be predicted to quantify the health status of the gas turbine exhaust temperature.

[0071]

[0072] The elements in the exhaust temperature similarity vector S are 0≤s j ≤1, the element 0≤c in the operating coefficient vector C j ≤1, then

[0073] h∈[0,1]

[0074] By further setting a threshold judgment rule for the h value, it is used to predict the healthy state of the gas turbine exhaust temperature.

[0075] The prediction method of the present invention converts the exhaust temperature distance vector into an exhaust temperature similarity vector through a kernel function, and designs a calculation formula for the health of data samples, thereby scientifically quantifying the health status of the gas turbine exhaust temperature based on the similarity calculation method. The method has been proven to be effective in practice and is easy to verify the design of algorithms and parameters to improve the accuracy and sensitivity of the algorithm.

[0076] In summary, from the above description, it can be seen that the above embodiments of the present invention achieve the following technical effects:

[0077] 1. After selecting a section of fault-free data with different operating conditions, the prediction method of the present invention selects key operating condition features to form a fault-free operating condition data set and a fault-free exhaust temperature data set, fully considering the impact of the operating conditions on the exhaust temperature. By collecting more fault-free data sets that cover different operating conditions as much as possible, the method has adaptive adjustment capabilities in predicting the exhaust temperature health under different stable operating conditions and weather conditions.

[0078] 2. The prediction method of the present invention normalizes the fault-free operating condition data set and normalizes the predicted operating condition vector using the recorded normalization model, thereby facilitating the calculation of the correlation between the predicted operating condition vector and the fault-free operating condition data set to obtain the operating condition coefficient vector for health correction.

[0079] 3. The prediction method of the present invention simulates the torsion of the temperature field, twists and sorts the elements of the exhaust temperature vector in sequence, obtains a new set of exhaust temperature vectors, and uses this method to obtain multiple sets of exhaust temperature vectors, and calculates the minimum value of the distance between them and each sample of the fault-free exhaust temperature data set. The prediction method of the present invention fully considers the torsion angle between the exhaust temperature distribution and the actual corresponding combustion chamber, so that the prediction method has adaptive ability to predict the exhaust temperature health of different combustion chamber designs.

[0080] 4. The prediction method of the present invention converts the exhaust temperature distance vector into an exhaust temperature similarity vector through a kernel function, and designs a calculation formula for the health of data samples, thereby scientifically quantifying the health status of the gas turbine exhaust temperature based on the similarity calculation method. The method has been proven to be effective in practice and is easy to verify the design of algorithms and parameters to improve the accuracy and sensitivity of the algorithm.

[0081] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for predicting the health status of gas turbine exhaust temperature, characterized in that: The steps include: Step 1: Collect data sets of gas turbine trouble-free operation under different stable operating conditions, including trouble-free exhaust temperature data sets T t and the fault-free condition data set W t ; Step 2: the fault-free working condition data set W t Carry out data processing; Step 3: Select the data set W corresponding to the fault-free working condition t For the same working condition characteristics, a set of data samples to be predicted are collected in real time, including the predicted working condition vector W p and the predicted exhaust temperature vector T p ; Step 4: Process the predicted operating condition vector Wp, and then calculate the fault-free operating condition data set W after data processing. t The correlation coefficient between them is obtained to obtain the operating condition coefficient vector C; Step 5: Calculate the fault-free exhaust temperature data set T t and the predicted exhaust temperature vector T D The distance between them is taken as the exhaust temperature distance vector D, and is converted into the exhaust temperature similarity vector S; Step six, based on the operating coefficient vector C and the exhaust temperature similarity vector S, calculate the health h of the data sample to be predicted to quantify the health state of the gas turbine exhaust temperature.

2. The prediction method according to claim 1, characterized in that: In step 5, the predicted exhaust temperature vector T p The elements in are twisted and sorted in clockwise or counterclockwise order to obtain a set of twisted exhaust temperature vectors, and then twisted multiple times in order to obtain multiple sets of exhaust temperature vectors T p_i .

3. The prediction method according to claim 2, characterized in that: In step 5, the fault-free exhaust temperature data set T t Each fault-free exhaust temperature vector T t_j Respectively with multiple groups of exhaust temperature vectors T p_i The minimum value of the distance is calculated to obtain the exhaust temperature distance vector D.

4. The prediction method according to claim 3, characterized in that: In the step five, the exhaust temperature distance vector D is converted into an exhaust temperature similarity vector S by constructing a kernel function, an exponential function or a logarithmic function.

5. The prediction method according to claim 4, characterized in that: In step six, each element in the operating condition coefficient vector C is multiplied by the corresponding element in the exhaust temperature similarity vector S, and the maximum value is taken as the health h of the data sample.

6. The prediction method according to claim 5, characterized in that: The exhaust temperature similarity vector S contains elements 0≤s j ≤1, the element 0≤c in the operating coefficient vector C j ≤1.

7. The prediction method according to any one of claims 2 to 5, characterized in that: The predicted exhaust temperature vector T p There are k elements in it, which are the exhaust temperature values ​​measured by k thermocouples distributed along the circumference. p Perform k-1 temperature twists to obtain k sets of exhaust temperature vectors T p_i .

8. The prediction method according to claim 1, characterized in that: In the step 2, the fault-free working condition data set W is analyzed according to the working condition characteristics. t Perform data normalization to obtain the normalized working condition data set W′ t , and the normalized model function γ.

9. The prediction method according to claim 8, characterized in that: In step 4, the predicted operating condition vector W is quantized using the normalized model function γ. p Perform data normalization to obtain the normalized working condition vector W′ p , and calculate W′ p and W′ t The correlation coefficient of the operating condition coefficient vector C is obtained.

10. The prediction method according to claim 8, characterized in that: In step 2, the normalization algorithm uses a minimum-maximum normalization or a decimal calibration normalization algorithm, so that the normalized working condition data set W′ t The values ​​of each operating condition feature in are distributed in the range of [0, 1].

11. The prediction method according to claim 9, characterized in that: In step 4, the normalized working condition vector W′ is calculated respectively p With the normalized working condition data set W′ t The absolute value of the Pearson correlation coefficient of each vector in is obtained to obtain the operating condition coefficient vector C.

12. The prediction method according to claim 1, characterized in that: The operating condition characteristics may be selected from: several or all of the following: rotation speed, load, IGV opening, compressor intake pressure, fuel temperature, fuel Weber index, ambient temperature, ambient atmospheric pressure, and ambient humidity.

Images