阅读说明：本技术 一种电站燃气轮机离线水洗时间的确定方法 (A kind of determination method of power station gas turbine off-line washing time ) 是由 杨永明 曲晓峰 张德阳 毛静轩 姜漫利 王兆光 苗东旭 翟俊鹏 傅磊 王克剑 于 2019-09-05 设计创作，主要内容包括：一种电站燃气轮机离线水洗时间的确定方法,属于电站燃气轮机离线水洗领域,本发明需要一种水洗时间合理、可以充分提高机组运行效率和设备利用率的电站燃气轮机离线水洗时间的确定方法。本发明中,从燃机电站数据库选取覆盖燃气轮机健康状态全工况集合A；对集合A稳态筛选得到运行数据集合B；建立压气机基准压比模型；利用集合C数据样本训练压气机基准压比模型；利用集合D数据样本计算压比残差；利用核密度估计法得到压比残差的上限值和下限值；将实时监测数据带入压气机基准压比模型得到压气机基准压比及其区间；计算实时压比降；设定水洗阈值；在对应的时间段内时进行离线水洗。本发明主要用于确定电站燃气轮机的离线水洗时间。(A kind of determination method of power station gas turbine off-line washing time, belong to power station gas turbine off-line washing field, the present invention need a kind of washing time rationally, can sufficiently improve unit operation efficiency and utilization rate of equipment and installations the power station gas turbine off-line washing time determination method.In the present invention, covering gas turbine health full working scope set A is chosen from combustion engine power station database；Set A stable state is screened to obtain operation data set B；It establishes compressor base pressure and compares model；Compare model using set C data sample training compressor base pressure；Pressure ratio residual error is calculated using set D data sample；The upper limit value and lower limit value of pressure ratio residual error are obtained using kernel density estimation method；It brings Real-time Monitoring Data into compressor base pressure and obtains compressor base pressure ratio and its section than model；Calculate real-time pressure ratio drop；Setting washing threshold value；Off-line washing is carried out when within the corresponding period.Present invention is mainly used for the off-line washing times for determining power station gas turbine.)

1. a kind of determination method of power station gas turbine off-line washing time, which is characterized in that it the following steps are included:

Step 1, the history health data that covering power station gas turbine full working scope is chosen from combustion engine power station database；That is set A；

Step 2 carries out stable state screening to set A, obtains the operation data under steady working condition, i.e. set B；It is collection by set B points Close C and set D；

Step 3, the boundary condition of analyzing influence compressor pressure ratio and factor determine the compressor base based on Support vector regression Quasi- pressure ratio seeks the input variable and output quantity of model, so that establishing the compressor base pressure based on Support vector regression compares mould Type；

Step 4 compares model using the state of health data sample training and test compressor base pressure of set C；

Data sample parameter is input to the compressor benchmark after training by step 5, the state of health data sample using set D In pressure ratio model, pressure ratio residual error is calculated；

It is step 6, for statistical analysis to the pressure ratio residual error using kernel density estimation method, and then when to obtain significance be α Pressure ratio residual error upper limit value e_hWith lower limit value e_l；

Step 7, real-time monitoring gas turbine obtain one group of duty parameter, bring compressor base pressure into than model, obtain compressor The predicted value of base pressure ratioThe upper limit in pressure ratio benchmark sectionThe lower limit in pressure ratio benchmark sectionAnd compressor Pressure ratio benchmark section

Step 8, the severity that power station gas turbine fouling is characterized using compressor pressure ratio drop calculate real-time pressure ratio drop；

Step 9 washes threshold value according to combustion engine group setup；

Step 10, when upper limit pressure ratio drop δ_hWhen reaching washing threshold value, the corresponding time is t_a, issue washing pre-warning signal；Work as lower limit δ drops in pressure ratio_lThe corresponding time is t when reaching washing threshold value_b, issue washing alarm signal；When in [t_a,t_b] in the period when into Row off-line washing.

2. the determination method of power station gas turbine off-line washing time according to claim 1 a kind of, which is characterized in that In In step 1, the history health data includes the temperature and pressure of compressor inlet, outlet, combustion engine power, environment temperature, IGV Valve opening.

3. the determination method of power station gas turbine off-line washing time according to claim 1 a kind of, which is characterized in that In In step 2, it is described to history health data carry out stable state screening the following steps are included:

Step 201: indicating stable state and transient, formula with a unified model are as follows:

In formula, x_tIndicate that the measured value of t moment power, μ indicate that the true value of power, m represent the rate of change of power；ε_tIt represents Random error of the power in t moment；

Step 202: the difference of the stable state and transient is whether load -response-speed is equal to 0；Using two neighboring The difference DELTA x of moment parameter is as characteristic statistic, formula are as follows:

Δ x=x_t-x_t-1=m+ (ε_t-ε_t-1)

Step 203: the characteristic statistic Δ x's is desired for m, according to the property of time series, in sample time window The mean value of characteristic statistic Δ x carries out desired estimation, formula are as follows:

In formula, h indicates to sentence the number of samples in steady time window, acquires a data, statistics from combustion engine power station database 1min The data of 20min are screened, therefore h=20；

Step 204: using the method for Estimating Confidence Interval,WithIt is to be determined by one section of long period steady state data in set B Two statistics, for give level of significance α, meet:

If m=0 is located atIn section, then there is the confidence level of (1- α) to think, system is in stable state in the period；Pass through Such method screens set A stable state at set B, is set C, set D two parts by set B points, enters step 3.

4. the determination method of power station gas turbine off-line washing time according to claim 1 a kind of, which is characterized in that In In step 5, the parameter includes combustion engine power, atmospheric temperature, atmospheric pressure, atmospheric humidity, compressor inlet temperature, compressor Inlet pressure and IGV aperture.

5. the determination method of power station gas turbine off-line washing time according to claim 1 a kind of, which is characterized in that In In step 5, the pressure ratio residual error is the difference of compressor pressure ratio actual observation value and Support vector regression model output value.

6. the determination method of power station gas turbine off-line washing time according to claim 1 a kind of, which is characterized in that In In step 8, when compressor base pressure ratio takes base pressure than the section upper limitWhen, calculating pressure ratio drop has:

When compressor base pressure ratio takes base pressure to compare interval limitWhen, calculating pressure ratio drop has:

In formula: δ_hFor compressor upper limit pressure ratio drop, δ_lFor compressor lower limit pressure ratio drop, π_cFor compressor actual observation pressure ratio,For Support vector regression model output value, i.e. compressor base pressure ratio, e_hFor the pressure ratio residual error upper limit, e_lFor pressure ratio residual error lower limit.

Technical field

The invention belongs to power station gas turbine off-line washing fields, and in particular to when a kind of power station gas turbine off-line washing Between determination method.

Background technique

Power station gas turbine needs constantly to suck at runtime air, may contain dust, dust, insect in air Deng, it is constantly deposited in compressor, pressure ratio and efficiency is caused to reduce, compressor is made to be easy to happen the serious accidents such as surge, reduction Safety, reliability and the economy of unit.Therefore offline water must be carried out to compressor by the aqueous solution containing detergent It washes, guarantees that combined cycle unit power and efficiency are in higher level.At present for the off-line washing of power station gas turbine, one As carried out according to the long short timer of its runing time, but this method does not fully consider the actual motion state of gas turbine, by It is influenced in being changed by external environment, the performance degradation of gas turbine can not may be made in periodically in the pair of certain factor Under, scaling rate increases, and performance degradation is accelerated.

In practical off-line washing matter of time, in order to preferably carry out off-line washing research, need to consider power station combustion gas The severity of turbine generation fouling failure.Modern production requires the overall efficiency for guaranteeing GTCC power plant to the greatest extent, existing Method for washing be as long as that gas turbine occurs flow passage component fouling failure and will take off-line washing measure immediately, but it is above-mentioned Method for washing cannot carry out purpose off-line washing according to the fault degree and development trend of fouling in the gas turbine operation of power station. And current method for washing only in some regular time point, and cannot in conjunction with power plant actual motion management, it is such from Line washing time is not scientific and reasonable, and the utilization rate of unit operation efficiency, equipment is low.

It is therefore desirable to which a kind of washing time is reasonable, can sufficiently improve the electricity of unit operation efficiency and utilization rate of equipment and installations It stands the determination method of gas turbine off-line washing time.

Summary of the invention

Unreasonable, unit operation efficiency, the utilization rate of equipment are low for existing method for washing washing time by the present invention Defect, provide a kind of washing time rationally, can sufficiently improve the power station combustion gas wheel of unit operation efficiency and utilization rate of equipment and installations The determination method of machine off-line washing time.

A kind of technical solution of the determination method of power station gas turbine off-line washing time according to the present invention is as follows:

The determination method of power station gas turbine off-line washing time according to the present invention a kind of, it the following steps are included:

Step 1, the history health data that covering power station gas turbine full working scope is chosen from combustion engine power station database；Gather A；

Step 2 carries out stable state screening to set A, obtains the operation data under steady working condition, i.e. set B；By set B points For set C and set D；

Step 3, the boundary condition of analyzing influence compressor pressure ratio and factor determine calming the anger based on Support vector regression Input variable and output quantity of the machine base pressure than seeking model, to establish the compressor base pressure based on Support vector regression Compare model；

Step 4 compares model using the state of health data sample training compressor base pressure of set C；

Data sample parameter is input to the compressor after training by step 5, the state of health data sample using set D Base pressure is than calculating pressure ratio residual error in model；

It is step 6, for statistical analysis to the pressure ratio residual error using kernel density estimation method, and then obtain significance and be The upper limit value e of pressure ratio residual error when α_hWith lower limit value e_l；

Step 7, real-time monitoring gas turbine obtain one group of duty parameter, bring compressor base pressure into than model, are pressed The predicted value of mechanism of qi base pressure ratioThe upper limit in pressure ratio benchmark sectionThe lower limit in pressure ratio benchmark sectionAnd pressure Mechanism of qi pressure ratio benchmark section

Step 8, the severity that power station gas turbine fouling is characterized using compressor pressure ratio drop, calculate real-time pressure ratio Drop；

Step 9 washes threshold value according to combustion engine group setup；

Step 10, when upper limit pressure ratio drop δ_hWhen reaching washing threshold value, the corresponding time is t_a, issue washing pre-warning signal；When δ drops in lower limit pressure ratio_lThe corresponding time is t when reaching washing threshold value_b, issue washing alarm signal；When in [t_a,t_b] in the period Shi Jinhang off-line washing.

Further: in step 1, the history health data includes the temperature and pressure of compressor inlet, outlet, combustion Machine power, environment temperature, IGV valve opening.

Further: in step 2, it is described to history health data carry out stable state screening the following steps are included:

Step 201: indicating stable state and transient, formula with a unified model are as follows:

In formula, x_tIndicate that the measured value of t moment power, μ indicate that the true value of power, m represent the rate of change of power；ε_t Power is represented in the random error of t moment；

Step 202: the difference of the stable state and transient is whether load -response-speed is equal to 0；Using adjacent The difference DELTA x of two moment parameters is as characteristic statistic, formula are as follows:

Δ x=x_t-x_t-1=m+ (ε_t-ε_t-1)

Step 203: the characteristic statistic Δ x's is desired for m, according to the property of time series, with sample time window The mean value of interior characteristic statistic Δ x carries out desired estimation, formula are as follows:

In formula, h indicates to sentence the number of samples in steady time window, acquires a data from combustion engine power station database 1min, The data of statistics 20min are screened, therefore h=20；

Step 204: using the method for Estimating Confidence Interval,WithIt is by one section of long period steady state data in set B Two determining statistics, for giving level of significance α, satisfaction:

If m=0 is located atIn section, then there is the confidence level of (1- α) to think, system is in stable state in the period； Set A stable state is screened into set B by such method, is set C, set D two parts by set B points, enters step 3.

Further: in steps of 5, the parameter includes combustion engine power, atmospheric temperature, atmospheric pressure, atmospheric humidity, pressure Mechanism of qi inlet temperature, compressor inlet pressure and IGV aperture.

Further: in steps of 5, the pressure ratio residual error is that compressor pressure ratio actual observation value and support vector machines are returned Return the difference of model output value.

Further: in step 8, when compressor base pressure ratio takes base pressure than the section upper limitWhen, calculate pressure Have than drop:

When compressor base pressure ratio takes base pressure to compare interval limitWhen, calculating pressure ratio drop has:

In formula: δ_hFor compressor upper limit pressure ratio drop, δ_lFor compressor lower limit pressure ratio drop, π_cFor compressor actual observation pressure ratio,For Support vector regression model output value, i.e. compressor base pressure ratio, e_hFor the pressure ratio residual error upper limit, e_lFor under pressure ratio residual error Limit.

A kind of beneficial effect of the determination method of power station gas turbine off-line washing time according to the present invention is:

The determination method of power station gas turbine off-line washing time according to the present invention a kind of, utilizes data mining, machine The technologies such as device study, statistical theory propose the method that practical view feelings washing time determines.Fully consider power station gas turbine The severity of fouling failure occurs for actual motion state, so that it is determined that the off-line washing time, evades incipient fault.The present invention Considering the variation of the boundary conditions such as power and environmental factor influences, and proposes that support vector machines combines the side of statistics with Density Estimator Method determines that compressor base pressure than model, and then obtains compressor base pressure under different operating conditions and compares section.Present invention additionally contemplates that The single off-line washing moment is unsatisfactory for for power plant's practical decision requirements in scene, on the basis of pressure ratio benchmark section, in conjunction with offline The empirical criteria of washing calculates pressure ratio and drops bound, and pressure ratio drop upper and lower limit is reached to when washing threshold value its corresponding two Time point determines the time interval of off-line washing.Such off-line washing time is more scientific and reasonable, can sufficiently improve unit The utilization rate of operational efficiency and equipment, and then increase the performance of enterprises and market core competitiveness.The present invention is according to power station combustion gas The severity of fouling failure occurs for turbine (fuel is natural gas), scientifically and rationally determines power station gas turbine off-line washing Time interval, to improve the economic benefit of gas turbine operation efficiency and enterprise.

Detailed description of the invention

Fig. 1 is a kind of flow diagram of the determination method of power station gas turbine off-line washing time；

Fig. 2 is the stable state Sample Filter figure of initial data；

Fig. 3 is that the compressor base pressure based on Support vector regression compares model schematic；

Fig. 4-a is C Set-dissection training sample actual value and regressand value comparative examples figure；

Fig. 4-b is C Set-dissection test sample actual value and regressand value comparative examples figure；

Fig. 5-a is D Set-dissection sample pressure ratio residual error exemplary diagram；

Fig. 5-b is pressure ratio residual error cuclear density distribution curve exemplary diagram.

Specific embodiment

Below with reference to embodiment, the following further describes the technical solution of the present invention, and however, it is not limited to this, all right Technical solution of the present invention is modified or replaced equivalently, and without departing from the spirit and scope of the technical solution of the present invention, should all be contained Lid is within the protection scope of the present invention.