CN105809304B - Method for analyzing correlation of production operation parameters of power plant and pollution control facility - Google Patents
Method for analyzing correlation of production operation parameters of power plant and pollution control facility Download PDFInfo
- Publication number
- CN105809304B CN105809304B CN201410840009.3A CN201410840009A CN105809304B CN 105809304 B CN105809304 B CN 105809304B CN 201410840009 A CN201410840009 A CN 201410840009A CN 105809304 B CN105809304 B CN 105809304B
- Authority
- CN
- China
- Prior art keywords
- data
- correlation coefficient
- correlation
- overrun
- parameter
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Abstract
The invention relates to the field of direct or distributed digital control systems for pollutants, in particular to a method for analyzing the correlation of production and operation parameters of a power plant and a pollution control facility. A method for the correlation analysis of production and operation parameters of a power plant and a pollution control facility is characterized by comprising the following steps: the system consists of three modules, which are respectively: the device comprises a noise reduction processing module, a long period correlation calculation module and a fine correlation calculation module. The method has the characteristics of high precision, low algorithm complexity, high operation speed and the like, and can meet the requirements of practical application.
Description
Technical Field
The invention relates to the field of direct or distributed digital control systems for pollutants, in particular to a method for analyzing the correlation of production and operation parameters of a power plant and a pollution control facility.
Background
At present, automatic monitoring of pollution sources is an advanced means for implementing environmental supervision, has the characteristics of automation, real time, on-line and the like, can provide massive sewage outlet monitoring data, enables an environmental protection department to master the latest pollution source emission and treatment facility operation conditions at the first time, and has incomparable advantages of the traditional environmental supervision and monitoring means. Since the concept of an automatic pollution source monitoring system is put forward in China, the investment of environmental protection departments at all levels in related fields is rapidly growing. Most of on-line monitoring systems used for environmental protection monitoring are used for monitoring result data (such as CEMS data and COD data) of pollution control facilities, namely, a transmission channel is established through various on-site acquisition means, the operation result of the enterprise environmental protection facilities is directly transmitted to an environmental protection department, and the traditional means is used for storing and analyzing. With the demands of economic development and environmental protection construction, the mode of 'end monitoring' is difficult to carry out effective monitoring and powerful supervision on pollution treatment facilities. The working condition online monitoring system can change the mode of online monitoring only the pollutant emission concentration and emission amount of enterprises, realizes the change from terminal monitoring to whole-process monitoring, improves the authenticity and accuracy of data, improves the scientificity and the notability of important pollution source monitoring data, provides powerful technical support for improving the environment supervision level, and plays a positive promoting role in timely discovering the environment illegal problem and strengthening the intensity of environment supervision and law enforcement. However, the current online monitoring has the following problems: firstly, the management system is not sound. At present, the design and product standard of the automatic pollution source monitoring device are yet to be further established and perfected. Secondly, the use and maintenance are not standard. Although many enterprises purchase the online monitoring device, the matching device and the auxiliary facilities cannot keep up with each other, and some enterprises even mix the pollution source automatic monitoring device with the chemical agent, so that the pollution source online monitoring equipment is corroded. Thirdly, the data reliability is not high. Few enterprises have only fear of exceeding pollution discharge and penalty, and move hands and feet on a display, so that data fluctuate within a certain range throughout the year, never exceed the standard, and are greatly different from the sampling monitoring data of environmental protection departments. With the continuous expansion of the application range of the data of the automatic pollution source monitoring facility, how to further strengthen the operation management of the automatic pollution source monitoring facility and ensure the operation stability and the data accuracy of the automatic pollution source monitoring system are more and more important.
The low data reliability is the most direct reason for restricting the application of the pollution source automatic monitoring system, and is specifically represented by data loss, data abnormity, data overrun, a constant value, data jumping, data validity abnormity, data correlation abnormity and the like, and the data correlation abnormity is the most obscure and difficult problem to find. The existing theory and method for correlation analysis are not comprehensive enough, and the practical production problem in the field of environmental protection cannot be solved.
Disclosure of Invention
The invention provides a method for determining correlation abnormal data in pollution source automatic monitoring data in order to overcome the defects of the prior art, and discloses a method for analyzing the correlation of production operation parameters of a power plant and a pollution control facility.
The invention achieves the purpose by the following technical scheme:
a method for the correlation analysis of production and operation parameters of a power plant and a pollution control facility is characterized by comprising the following steps: the system consists of three modules, which are respectively: a noise reduction processing module, a long period correlation calculation module and a fine correlation calculation module,
a noise reduction processing module: the noise reduction processing of the acquired time sequence data comprises deleting unavailable data, supplementing missing data and smoothing the time sequence data, and the noise reduction processing module ensures that production operation data can be used for correlation analysis and retains the original main characteristics of the data;
a long period correlation calculation module: performing correlation analysis on the acquired data in a full time period, calculating correlation coefficients among parameter indexes of the time series data according to a Pearson correlation coefficient formula, and checking whether the correlation coefficients among the parameter indexes of the time series data are within a preset correlation coefficient range, so as to determine that the correlation coefficients exceed the limit parameters on a large trend;
a refinement relevance calculation module: after the overrun parameter is determined, scanning data segments with relatively large fluctuation amplitude, sorting and combining the large-amplitude fluctuation data segments, and determining whether the correlation of the overrun parameter is abnormal or not by a multi-item correlation method.
The method for analyzing the correlation of the production operation parameters of the power plant and the pollution control facility is characterized by comprising the following steps of: correlation analysis involves two types of data: the model parameters and the time-series data,
the model parameters comprise unit load, coal burning quantity, total unit air supply quantity, inlet flue gas flow, outlet flue gas flow, booster fan current and induced draft fan current;
time series data is real-time data taken from power plant and pollution treatment facility production operations, each data point having a time tag.
The method for analyzing the correlation of the production operation parameters of the power plant and the pollution control facility is characterized by comprising the following steps of: the noise reduction processing module comprises a jump checking module, a purging checking module and a data noise reduction module:
a jump checking module: checking the jumping intensity of the read time sequence data, if the time sequence data in a period has frequent jumping intensity, the data in the period can not be analyzed for relativity, and directly deleting the data in the period;
a purging inspection module: checking the read time sequence data, performing purging filtration for the phenomenon that data jump happens occasionally in a period of time, namely purging the jump data, and performing linear interpolation operation to compensate the missing data;
a data noise reduction module: checking the jumping intensity of the read time sequence data, if the time sequence data in a period has frequent jumping with a small degree, adopting an arithmetic moving average algorithm to perform noise reduction processing on the data in the period, and using the data subjected to the noise reduction processing for correlation analysis.
The method for analyzing the correlation of the production operation parameters of the power plant and the pollution control facility is characterized by comprising the following steps of: the long period correlation calculation module comprises:
pearson correlation coefficient matrix: forming a Pearson correlation coefficient matrix by the Pearson correlation coefficient groups between every two model parameters, wherein the Pearson correlation coefficient matrix is a symmetric matrix;
presetting a correlation coefficient range: and through statistical analysis of the normal data samples, obtaining the upper and lower limits of multiple standard deviations of the mean value of the Pearson correlation coefficient as the preset correlation coefficient range.
The method for analyzing the correlation of the production operation parameters of the power plant and the pollution control facility is characterized by comprising the following steps of: the refinement relevance calculation module comprises:
an integral correlation coefficient error rate calculation module: calculating integral correlation coefficient of the overrun parameter of the current data, comparing the integral correlation coefficient with standard integral correlation coefficient calculated by an integral correlation coefficient self-learning model, and calculating integral correlation coefficient error rate;
and the pilson correlation coefficient of the overrun parameter overrun proportion calculation module again: and calculating a Pearson correlation coefficient matrix between the model parameters of the large-amplitude fluctuation section, counting the number of the secondary overrun of the Pearson correlation coefficients of the overrun parameters, and calculating the secondary overrun occupation ratio of the Pearson correlation coefficients of the overrun parameters.
An abnormality determination module: and determining an abnormal judgment condition according to the error rate of the integral correlation coefficient and the proportion of the pilson correlation coefficient of the overrun parameter exceeding the limit again, thereby judging whether the data of the overrun parameter has correlation abnormality.
The method for analyzing the correlation of the production operation parameters of the power plant and the pollution control facility is characterized by comprising the following steps of: the multiple standard deviation selection in the range of the preset correlation coefficient is calculated and adjusted according to the experimental result, and generally 3 standard deviation ranges are selected.
The method for analyzing the correlation of the production operation parameters of the power plant and the pollution control facility is characterized by comprising the following steps of: the calculation of the integral correlation coefficient is based on the selection of main parameters, the main parameters are the most important parameters in the production operation of power plants and pollution control facilities, and the invention selects the unit load.
The method for analyzing the correlation of the production operation parameters of the power plant and the pollution control facility is characterized by comprising the following steps of: the abnormality determination condition includes two conditions, one of which is satisfied, that is, the overrun parameter is determined to be abnormal: firstly, the error rate of the integral correlation coefficient exceeds 10%, and the overrun percentage of the Pearson correlation coefficient of the overrun parameter exceeds 30%; secondly, the error rate of the integral correlation coefficient exceeds 5 percent, and the overrun percentage of the Pearson correlation coefficient of the overrun parameter exceeds 70 percent.
The method for analyzing the correlation of the production operation parameters of the power plant and the pollution control facility is characterized by comprising the following steps of: the method is implemented in sequence according to the following steps:
a. initializing a system, starting a model and a real-time database, and reading in model parameters and time sequence data;
b. denoising the acquired time sequence data, namely deleting unavailable data, supplementing missing data and smoothing denoising, ensuring that production operation data can be used for correlation analysis through a denoising module, and keeping the original main characteristics of the data;
c. calculating a correlation coefficient between parameter indexes of the real-time data according to a Pearson correlation coefficient formula;
d. through statistical analysis on historical data, determining a preset range of long-period strong correlation coefficient data of parameter indexes, and checking whether correlation coefficients among the parameter indexes of real-time data are within the preset correlation coefficient range;
e. capturing data segments with the fluctuation of the correlation coefficient exceeding a preset range, sorting and combining, and if the correlation coefficient is within the preset range, ending the model;
f. and calculating a data correlation matrix of the fluctuation section, and judging whether the data correlation of the overrun parameter is abnormal or not by combining the error rate of the integral correlation coefficient and the proportion of the overrun parameter to the Pearson correlation coefficient again through refined correlation calculation.
The model refers to a method for analyzing the correlation of production and operation parameters of a power plant and a pollution control facility, and the model parameters comprise: unit load, coal burning quantity, total unit air supply quantity, inlet flue gas flow, outlet flue gas flow, booster fan current and draught fan current. The relationship between the parameter indexes is as follows:
1. load-coal combustion amount: positive correlation (load becomes large, coal combustion amount becomes large);
2. load-total air supply of the unit: positive correlation;
3. load-inlet flue gas flow: positive correlation;
4. load-booster fan current: positive correlation;
5. load-current of the induced draft fan: positive correlation;
6. coal-fired quantity-total air supply of unit: positive correlation;
7. coal combustion amount-inlet flue gas flow: positive correlation;
8. coal burning amount-booster fan current: positive correlation;
9. coal burning quantity-current of a draught fan: positive correlation;
10. total air supply to the unit-inlet flue gas flow: positive correlation;
11. total air supply of the unit-current of the booster fan: positive correlation;
12. total air output of the unit-current of the induced draft fan: positive correlation;
13. inlet flue gas flow-outlet flue gas flow: positive correlation is carried out, and the outlet flue gas flow is larger than the inlet flue gas flow;
14. inlet flue gas flow-booster fan current: positive correlation;
15. inlet flue gas flow-induced draft fan current: positive correlation;
16. booster fan current-fan current: and (4) positively correlating.
The invention provides a method for the correlation analysis of production operation parameters of a power plant and a pollution control facility. The method has the characteristics of high precision, low algorithm complexity, high operation speed and the like on the premise of ensuring the correctness of the data result, and can meet the requirements of practical application.
Drawings
FIG. 1 is a flow chart of a method for correlation analysis of production operating parameters of a power plant and a pollution control facility according to the present invention;
FIG. 2 is a schematic diagram of a method for calculating an integral correlation coefficient according to the present invention.
Detailed Description
The invention is further illustrated by the following specific examples.
Example 1
A method for the correlation analysis of production and operation parameters of a power plant and a pollution control facility comprises three modules, which are respectively as follows: a noise reduction processing module, a long period correlation calculation module and a fine correlation calculation module,
a noise reduction processing module: the noise reduction processing of the acquired time sequence data comprises deleting unavailable data, supplementing missing data and smoothing the time sequence data, and the noise reduction processing module ensures that production operation data can be used for correlation analysis and retains the original main characteristics of the data;
in this embodiment, the noise reduction processing module includes a jump checking module, a purge checking module, and a data noise reduction module:
a jump checking module: checking the jumping intensity of the read time sequence data, if the time sequence data in a period has frequent jumping intensity, the data in the period can not be analyzed for relativity, and directly deleting the data in the period;
a purging inspection module: checking the read time sequence data, performing purging filtration for the phenomenon that data jump happens occasionally in a period of time, namely purging the jump data, and performing linear interpolation operation to compensate the missing data;
a data noise reduction module: checking the jumping intensity of the read time sequence data, if the time sequence data in a period has frequent jumping with a small degree, adopting an arithmetic moving average algorithm to perform noise reduction processing on the data in the period, and using the data subjected to the noise reduction processing for correlation analysis.
A long period correlation calculation module: performing correlation analysis on the acquired data in a full time period, calculating correlation coefficients among parameter indexes of the time series data according to a Pearson correlation coefficient formula, and checking whether the correlation coefficients among the parameter indexes of the time series data are within a preset correlation coefficient range, so as to determine that the correlation coefficients exceed the limit parameters on a large trend;
in this embodiment, the long-period correlation calculation module includes:
pearson correlation coefficient matrix: forming a Pearson correlation coefficient matrix by the Pearson correlation coefficient groups between every two model parameters, wherein the Pearson correlation coefficient matrix is a symmetric matrix;
presetting a correlation coefficient range: and through statistical analysis of the normal data samples, obtaining the upper and lower limits of multiple standard deviations of the mean value of the Pearson correlation coefficient as the preset correlation coefficient range.
A refinement relevance calculation module: after the overrun parameter is determined, scanning data segments with relatively large fluctuation amplitude, sorting and combining the large-amplitude fluctuation data segments, and determining whether the correlation of the overrun parameter is abnormal or not by a multi-item correlation method.
In this embodiment, the refinement correlation calculation module includes:
an integral correlation coefficient error rate calculation module: calculating integral correlation coefficient of the overrun parameter of the current data, comparing the integral correlation coefficient with standard integral correlation coefficient calculated by an integral correlation coefficient self-learning model, and calculating integral correlation coefficient error rate;
and the pilson correlation coefficient of the overrun parameter overrun proportion calculation module again: calculating a Pearson correlation coefficient matrix between model parameters of the large-amplitude fluctuation section, counting the number of the secondary overrun of the Pearson correlation coefficients of the overrun parameters, and calculating the ratio of the secondary overrun of the Pearson correlation coefficients of the overrun parameters;
an abnormality determination module: and determining an abnormal judgment condition according to the error rate of the integral correlation coefficient and the proportion of the pilson correlation coefficient of the overrun parameter exceeding the limit again, thereby judging whether the data of the overrun parameter has correlation abnormality.
The method for analyzing the correlation of the production operation parameters of the power plant and the pollution control facility is characterized by comprising the following steps of: correlation analysis involves two types of data: the model parameters and the time-series data,
the model parameters comprise unit load, coal burning quantity, total unit air supply quantity, inlet flue gas flow, outlet flue gas flow, booster fan current and induced draft fan current;
time series data is real-time data taken from power plant and pollution treatment facility production operations, each data point having a time tag.
The multiple standard deviation selection in the range of the preset correlation coefficient is calculated and adjusted according to the experimental result, and generally 3 standard deviation ranges are selected.
The integral correlation coefficient is calculated according to the selection of main parameters, the main parameters are the most important parameters in the production operation of a power plant and a pollution control facility, and the unit load is selected in the embodiment.
The abnormality determination condition includes two conditions, one of which is satisfied, that is, the overrun parameter is determined to be abnormal: firstly, the error rate of the integral correlation coefficient exceeds 10%, and the overrun percentage of the Pearson correlation coefficient of the overrun parameter exceeds 30%; secondly, the error rate of the integral correlation coefficient exceeds 5 percent, and the overrun percentage of the Pearson correlation coefficient of the overrun parameter exceeds 70 percent.
The method is implemented by the following steps in sequence:
a. initializing a system, starting a model and a real-time database, and reading in model parameters and time sequence data;
b. denoising the acquired time sequence data, namely deleting unavailable data, supplementing missing data and smoothing denoising, ensuring that production operation data can be used for correlation analysis through a denoising module, and keeping the original main characteristics of the data;
c. calculating a correlation coefficient between parameter indexes of the real-time data according to a Pearson correlation coefficient formula;
d. through statistical analysis on historical data, determining a preset range of long-period strong correlation coefficient data of parameter indexes, and checking whether correlation coefficients among the parameter indexes of real-time data are within the preset correlation coefficient range;
e. capturing data segments with the fluctuation of the correlation coefficient exceeding a preset range, sorting and combining, and if the correlation coefficient is within the preset range, ending the model;
f. and calculating a data correlation matrix of the fluctuation section, and judging whether the data correlation of the overrun parameter is abnormal or not by combining the error rate of the integral correlation coefficient and the proportion of the overrun parameter to the Pearson correlation coefficient again through refined correlation calculation.
The model refers to a method for analyzing the correlation of production and operation parameters of a power plant and a pollution control facility, and the model parameters comprise: unit load, coal burning quantity, total unit air supply quantity, inlet flue gas flow, outlet flue gas flow, booster fan current and draught fan current. The relationship between the parameter indexes is as follows:
1. load-coal combustion amount: positive correlation (load becomes large, coal combustion amount becomes large);
2. load-total air supply of the unit: positive correlation;
3. load-inlet flue gas flow: positive correlation;
4. load-booster fan current: positive correlation;
5. load-current of the induced draft fan: positive correlation;
6. coal-fired quantity-total air supply of unit: positive correlation;
7. coal combustion amount-inlet flue gas flow: positive correlation;
8. coal burning amount-booster fan current: positive correlation;
9. coal burning quantity-current of a draught fan: positive correlation;
10. total air supply to the unit-inlet flue gas flow: positive correlation;
11. total air supply of the unit-current of the booster fan: positive correlation;
12. total air output of the unit-current of the induced draft fan: positive correlation;
13. inlet flue gas flow-outlet flue gas flow: positive correlation is carried out, and the outlet flue gas flow is larger than the inlet flue gas flow;
14. inlet flue gas flow-booster fan current: positive correlation;
15. inlet flue gas flow-induced draft fan current: positive correlation;
16. booster fan current-fan current: and (4) positively correlating.
Referring now to fig. 1 and 2, the method for correlation analysis of production operation parameters of power plants and pollution control facilities according to the present invention is specifically described:
step S201, initialization phase. The invention relates to two types of basic data, one is a power plant and pollution control facility production operation model parameter, the other is real-time data generated in the power plant and pollution control facility production operation, and each data has a time attribute and a unique parameter attribute. In this step, the model parameter { P } is selected1，P2，P3，…，PmStopping current historical data in a period of time, such as a parameter P acquired according to a sampling frequency of once per second1Time series data { x1，x2，x3，…，xn}。
Step S202, setting a jump test period according to the sampling frequency of the time series data, and calculating the ratio of the fluctuation degree index and the average fluctuation to the average ratio in the jump test period, wherein the calculation formula is as follows:
F＝|xmax-xmin|，
in the formula, xiAre data points;
n is the number of dots;
f is the amplitude of the dot;
f is the average amplitude of the point, and the data is divided into a plurality of sections with the length of m points during calculation, and the average amplitude is calculated;
when index >0.05 and ratio > 20%, step S203 is entered, otherwise, step S204 is entered.
And step S203, alarming the measuring point, prompting that the data is seriously jumped, and deleting all time sequence data in the jump checking period from the sample.
And S204, carrying out purging jump test, adopting 1hour calculation once, taking the data segment of the current hour and the first 1hour as data samples each time, carrying out variance calculation of 2hour data, preliminarily setting the data small period to be 5min, carrying out mean value calculation of the first 5min every 1min, then setting the variance range, and if the data point or the data segment exceeds the set variance range, considering that the data is purged and jumped, and further processing. And if the data jump length is larger than 1min, the step S205 is entered, otherwise, the step S206 is entered.
And S205, eliminating data in a time period when the jump length is greater than 1min, performing difference operation on missing data, and supplementing time sequence data.
And step S206, for the time sequence data segment with the jump length less than 1min, directly acquiring the data and removing the data.
Step S207, obtaining the current data, calculating the arithmetic moving average of each group of time series data, and replacing the current data, thereby obtaining a new group of data subjected to noise reduction processing. The new data is smoother and trend consistency is guaranteed compared to the previous data.
Step S208, after the noise reduction processing, the Pearson correlation coefficient r is calculated, the Pearson correlation coefficient between two parameters is calculated, and each parameter and other parameters are calculated. A pearson correlation coefficient matrix is then established between the parameters, as shown in table 1:
table 1:
the correlation coefficient r is calculated by the formula:
wherein x and y represent model parameters, rxy∈[-1,1]Representing the pearson correlation coefficient between parameter x and parameter y.
In table 1: r is12Pearson's correlation coefficient, r, for the column 1 heading column "unit load" and the row 2 heading column "coal burn amountijThe correlation coefficients of the pilson in the ith column title bar and the jth row title bar are obtained, and the like;
FGD (i.e., flow gas desulfurization) represents a Flue gas desulfurization unit;
all selected model parameters exhibit strong positive correlation.
Step S209, at first, another normal data sample is selected for statistics, and the upper and lower limits of the calculated average value 3sigma of the pearson correlation coefficient of the normal sample are used as the preset range of the pearson correlation coefficient. The production operation parameters of the power plant and the pollution control facility selected by the invention are in strong positive correlation, and the preset range is calculated to be R epsilon (0.67, 0.995). Verifying whether the value of the Pearson correlation coefficient matrix is within a preset range, if so, judging that the correlation of the parameters of all the data is normal, and performing step S217; otherwise, recording the overrun parameter Pi(i>0) Step S210 is performed.
Step S210, performing fluctuation scanning on the time series data, calculating a data segment with relatively large fluctuation of the data, intercepting, sorting and combining the data segments with large fluctuation, and performing further correlation analysis.
Step S211, a Pearson correlation coefficient matrix between the parameters of the large fluctuation section is calculated.
Step S212, calculating an overrun parameter P, wherein the preset range of the large-amplitude fluctuation section is wider and is recorded as R' epsilon (0.6,1) relative to the preset range R of the Pearson correlation coefficient of the data subjected to noise reduction processingiCalculating the overrun parameter P when the Pearson coefficient exceeds the R' number k in the large fluctuation section with other parametersiAgain the pearson correlation coefficient of (a) is over-limited by k/(m-1), λ ∈ [0,1 ∈]Where m is the number of model parameters.
Step S213, firstly, selecting the unit load as a main parameter, and calculating the time integral alpha of the data of the main parameter in a large-amplitude fluctuation section; secondly, acquiring an overrun parameter PiCalculate PiThe data of (b) is integrated over time over a period of large fluctuations. Finally, calculating the ratio delta of the integral alpha and the integral beta, wherein delta is the overrun parameter PiAnd an integral correlation coefficient between the main parameters.
Step S214, a plurality of normal data samples are additionally selected, and an overrun parameter P is calculatediAnd continuously correcting the fitting formula through the self-learning model.
Step S215, calculating an integral correlation coefficient when the time integral value of the main parameter is alpha according to the fitting formula F, and recording the result as delta', wherein the overrun parameter PiAnd the error rate (ξ) of the integral correlation coefficient between the main parameter is calculated by the formula:
step S216, the Clarson correlation coefficient of the overrun parameter is checked to overrun the lambda range and the error rate xi range of the integral correlation coefficient again, and the abnormity judgment is carried out, wherein the judgment conditions are as follows:
1. the error rate of the integral correlation coefficient exceeds 10% of xi, and the proportion of the Pearson correlation coefficient of the overrun parameter exceeds 30% of lambda again;
2. the error rate of the integrated correlation coefficient exceeds 5% of ξ and the pearson correlation coefficient of the overrun parameter again overrun by more than 70% of λ.
It is determined that either of the two conditions is satisfied: the overrun parameter PiData dependency anomaly of (2); otherwise, the overrun parameter PiThe data correlation of (2) is normal.
Step S217, the parameter correlation analysis is ended.
Claims (7)
1. A method for the correlation analysis of production and operation parameters of a power plant and a pollution control facility is characterized by comprising the following steps: the method comprises the following three steps: noise reduction processing, long period correlation calculation and refined correlation calculation,
and (3) noise reduction treatment: the noise reduction processing of the acquired time sequence data comprises deleting unavailable data, supplementing missing data and smoothing the time sequence data, and through the noise reduction processing, the production operation data can be used for correlation analysis, and original main characteristics of the data are reserved;
long period correlation calculation: performing correlation analysis on the acquired data in a full time period, calculating correlation coefficients among parameter indexes of the time series data according to a Pearson correlation coefficient formula, and checking whether the correlation coefficients among the parameter indexes of the time series data are within a preset correlation coefficient range, so as to determine that the correlation coefficients exceed the limit parameters on a large trend;
and (3) refining correlation calculation: after the overrun parameters are determined, scanning data segments with larger relative fluctuation amplitude, namely data segments exceeding 3 standard deviation ranges, sorting and combining the large-amplitude fluctuation data segments, and performing refined correlation calculation to jointly determine whether the correlation of the overrun parameters is abnormal or not;
the refinement relevance computation includes:
integral correlation coefficient error rate calculation: calculating integral correlation coefficient of the overrun parameter of the current data, comparing the integral correlation coefficient with standard integral correlation coefficient calculated by an integral correlation coefficient self-learning model, and calculating integral correlation coefficient error rate;
and (3) calculating the ratio of the Pearson correlation coefficient of the overrun parameter to the overrun ratio again: calculating a Pearson correlation coefficient matrix between model parameters of the large-amplitude fluctuation section, counting the number of the secondary overrun of the Pearson correlation coefficients of the overrun parameters, and calculating the ratio of the secondary overrun of the Pearson correlation coefficients of the overrun parameters;
and (3) judging the abnormality: and determining an abnormal judgment condition according to the error rate of the integral correlation coefficient and the proportion of the pilson correlation coefficient of the overrun parameter exceeding the limit again, thereby judging whether the data of the overrun parameter has correlation abnormality.
2. The method of correlation analysis of production operating parameters of power plants and pollution control facilities of claim 1, wherein: correlation analysis involves two types of data: the model parameters and the time-series data,
the model parameters comprise unit load, coal burning quantity, total unit air supply quantity, inlet flue gas flow, outlet flue gas flow, booster fan current and induced draft fan current;
time series data is real-time data taken from power plant and pollution treatment facility production operations, each data point having a time tag.
3. The method of correlation analysis of production operating parameters of power plants and pollution control facilities of claim 1, wherein: the noise reduction processing comprises jump checking, purging checking and data noise reduction:
jump checking: checking the jumping intensity of the read time sequence data, if the time sequence data in a period has frequent jumping intensity, the data in the period can not be analyzed for relativity, and directly deleting the data in the period;
and (4) purging and checking: checking the read time sequence data, performing purging filtration for the phenomenon that data jump happens occasionally in a period of time, namely purging the jump data, and performing linear interpolation operation to compensate the missing data;
data denoising: checking the jumping intensity of the read time sequence data, if the time sequence data in a period has frequent jumping with a relatively low degree, performing noise reduction processing on the data in the period by adopting an arithmetic moving average algorithm, and using the data subjected to the noise reduction processing for correlation analysis;
setting a jump test period according to the sampling frequency of the time series data, and calculating the ratio of the fluctuation degree index and the average fluctuation to the average ratio in the jump test period, wherein the calculation formula is as follows:
F＝|xmax-xmin|，
in the formula, xiAre data points;
n is the number of dots;
f is the amplitude of the dot;
when index is less than or equal to 0.05 or ratio is less than or equal to 20%, it is considered that the time-series data has occurred with less frequent jitter.
4. The method of correlation analysis of production operating parameters of power plants and pollution control facilities of claim 1, wherein: the long period correlation calculation comprises the following steps:
pearson correlation coefficient matrix: forming a Pearson correlation coefficient matrix by the Pearson correlation coefficient groups between every two model parameters, wherein the Pearson correlation coefficient matrix is a symmetric matrix;
presetting a correlation coefficient range: through statistical analysis of the normal data samples, a range between an upper limit and a lower limit of three times of standard deviation of the Pearson correlation coefficient mean is obtained as a preset correlation coefficient range.
5. The method of correlation analysis of production operating parameters of power plants and pollution control facilities of claim 1, wherein: and calculating the integral correlation coefficient according to the selection of the main parameter, wherein the main parameter selects the unit load.
6. The method of correlation analysis of production operating parameters of power plants and pollution control facilities of claim 1, wherein: the abnormality determination condition includes two conditions, one of which is satisfied, that is, the overrun parameter is determined to be abnormal: firstly, the error rate of the integral correlation coefficient exceeds 10%, and the overrun percentage of the Pearson correlation coefficient of the overrun parameter exceeds 30%; secondly, the error rate of the integral correlation coefficient exceeds 5 percent, and the overrun percentage of the Pearson correlation coefficient of the overrun parameter exceeds 70 percent.
7. The method for correlation analysis of production operation parameters of power plants and pollution control facilities according to any one of claims 1 to 6, wherein: the method is implemented in sequence according to the following steps:
a. initializing a system, starting a model and a real-time database, and reading in model parameters and time sequence data;
b. denoising the acquired time sequence data, namely deleting unavailable data, supplementing missing data and smoothing denoising, ensuring that production operation data can be used for correlation analysis through denoising, and keeping the original main characteristics of the data;
c. calculating a correlation coefficient between parameter indexes of the real-time data according to a Pearson correlation coefficient formula;
d. through statistical analysis on historical data, determining a preset range of long-period strong correlation coefficient data of parameter indexes, and checking whether correlation coefficients among the parameter indexes of real-time data are within the preset correlation coefficient range;
e. capturing data segments with the fluctuation of the correlation coefficient exceeding a preset range, and sorting and combining the data segments for further correlation analysis, wherein the further correlation analysis comprises the following steps S211-S217:
step S211, calculating a Pearson correlation coefficient matrix between parameters of the large-amplitude fluctuation section;
step S212, calculating an overrun parameter P, wherein the preset range of the large-amplitude fluctuation section is wider and is recorded as R' epsilon (0.6,1) relative to the preset range R of the Pearson correlation coefficient of the data subjected to noise reduction processingiCalculating the overrun parameter P when the Pearson coefficient exceeds the R' number k in the large fluctuation section with other parametersiAgain the pearson correlation coefficient of (a) is over-limited by k/(m-1), λ ∈ [0,1 ∈]Wherein m is the number of model parameters;
step S213, firstly, selecting the unit load as a main parameter, and calculating the time integral alpha of the data of the main parameter in a large-amplitude fluctuation section; secondly, acquiring an overrun parameter PiCalculate PiThe data is integrated with the time at a large-amplitude fluctuation section, and finally, the ratio delta of the integral alpha to the integral beta is calculated, wherein the delta is the overrun parameter PiAnd the integral correlation coefficient between the main parameters;
step S214, a plurality of normal data samples are additionally selected, and an overrun parameter P is calculatediThe fitting formula F between the time integral and the main parameter and the time integral is continuously corrected through a self-learning model;
step S215, calculating an integral correlation coefficient when the time integral value of the main parameter is alpha according to the fitting formula F, and recording the result as delta', wherein the overrun parameter PiAnd the error rate (ξ) of the integral correlation coefficient between the main parameter is calculated by the formula:
step S216, the Clarson correlation coefficient of the overrun parameter is checked to overrun the lambda range and the error rate xi range of the integral correlation coefficient again, and the abnormity judgment is carried out, wherein the judgment conditions are as follows:
1. the error rate of the integral correlation coefficient exceeds 10% of xi, and the proportion of the Pearson correlation coefficient of the overrun parameter exceeds 30% of lambda again;
2. the error rate of the integral correlation coefficient exceeds 5% of xi, and the proportion of the Pearson correlation coefficient of the overrun parameter exceeds 70% of lambda again;
it is determined that either of the two conditions is satisfied: the overrun parameter PiData dependency anomaly of (2); otherwise, the overrun parameter PiThe data correlation of (2) is normal;
step S217, the parameter correlation analysis is ended.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410840009.3A CN105809304B (en) | 2014-12-27 | 2014-12-27 | Method for analyzing correlation of production operation parameters of power plant and pollution control facility |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410840009.3A CN105809304B (en) | 2014-12-27 | 2014-12-27 | Method for analyzing correlation of production operation parameters of power plant and pollution control facility |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105809304A CN105809304A (en) | 2016-07-27 |
| CN105809304B true CN105809304B (en) | 2021-04-09 |
Family
ID=56979982
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410840009.3A Active CN105809304B (en) | 2014-12-27 | 2014-12-27 | Method for analyzing correlation of production operation parameters of power plant and pollution control facility |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105809304B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110895404A (en) * | 2018-09-12 | 2020-03-20 | 长鑫存储技术有限公司 | Method and system for automatically detecting correlation between integrated circuit parameters |
| CN110469522A (en) * | 2019-08-13 | 2019-11-19 | 浪潮通用软件有限公司 | A kind of method for detecting abnormality and device of drainage system |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8301464B1 (en) * | 2008-07-18 | 2012-10-30 | Cave Consulting Group, Inc. | Method and system for producing statistical analysis of medical care information |
| CN103207944A (en) * | 2013-02-04 | 2013-07-17 | 国家电网公司 | Electric power statistical index relevance analysis method |
| CN103678869A (en) * | 2013-09-17 | 2014-03-26 | 中国人民解放军海军航空工程学院青岛校区 | Prediction and estimation method of flight parameter missing data |
| CN103813355A (en) * | 2014-02-21 | 2014-05-21 | 厦门大学 | Identification method for anomalous points of cooperative synchronization in distributed network |
-
2014
- 2014-12-27 CN CN201410840009.3A patent/CN105809304B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8301464B1 (en) * | 2008-07-18 | 2012-10-30 | Cave Consulting Group, Inc. | Method and system for producing statistical analysis of medical care information |
| CN103207944A (en) * | 2013-02-04 | 2013-07-17 | 国家电网公司 | Electric power statistical index relevance analysis method |
| CN103678869A (en) * | 2013-09-17 | 2014-03-26 | 中国人民解放军海军航空工程学院青岛校区 | Prediction and estimation method of flight parameter missing data |
| CN103813355A (en) * | 2014-02-21 | 2014-05-21 | 厦门大学 | Identification method for anomalous points of cooperative synchronization in distributed network |
Non-Patent Citations (2)
| Title |
|---|
| A Correlative Analysis Process in a Visual Analytics Environment;Abish Malik etc.;《2012 IEEE Conference on Visual Analytics Science and Technology (VAST)》;20121019;第33-42页 * |
| 应用皮尔逊相关系数算法查找异常电能表用户;王涓 等;《电力需求侧管理》;20140331;第16卷(第2期);第52-54页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105809304A (en) | 2016-07-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109493250B (en) | Method for evaluating denitration capability of SCR reactor | |
| Rieger et al. | Progress in sensor technology-progress in process control? Part I: Sensor property investigation and classification | |
| CN110094251B (en) | SCR catalyst performance degradation analysis method based on time-interval multi-model modeling | |
| CN112308273A (en) | Memory, petrochemical enterprise pollution discharge management method, device and equipment | |
| CN111898691A (en) | River sudden water pollution early warning tracing method, system, terminal and medium | |
| CN112034800B (en) | Method, system, medium and terminal for calculating unorganized emission of volatile organic pollutants | |
| CN112232571B (en) | Method for predicting concentration of main pollutants in waste gas | |
| CN102436232B (en) | Equipment maintaining system for factories for purifying natural gas with high content of sulfur | |
| CH702625A2 (en) | System and method for monitoring a gas turbine. | |
| He et al. | Environmental tax, polluting plants’ strategies and effectiveness: Evidence from China | |
| CN112034801B (en) | Method, system and terminal for calculating pollution discharge coefficient of total amount of volatile organic pollutants | |
| CN105809304B (en) | Method for analyzing correlation of production operation parameters of power plant and pollution control facility | |
| CN114757380A (en) | Thermal power plant fault early warning system and method, electronic equipment and storage medium | |
| CN105808902B (en) | Qualitative method for analyzing operation condition of wet desulphurization system | |
| JP2008112428A (en) | Method and apparatus for statistically predicting quality of inflow water in water disposal facility | |
| CN116263850A (en) | Online sewage water quality early warning method combining offline simulation data | |
| CN116432439A (en) | Urban river sewage receiving capacity planning method and system based on numerical simulation | |
| CN108491995B (en) | Key control factor screening method for drinking water risk factor identification | |
| RU2295590C1 (en) | Method of the statistical control over the quality of the electrode products | |
| CN110515796B (en) | Cortex learning-based anomaly detection method and device and terminal equipment | |
| CN113313529A (en) | Finished oil sales amount prediction method based on time regression sequence | |
| CN116976562A (en) | Fixed point source pollutant and carbon emission list dynamic accounting system and method | |
| CN115906687B (en) | Quantitative analysis and evaluation method for river environmental influence of industrial life water intake and water withdrawal | |
| CN114742300B (en) | Denitration device boiler flue resistance early warning method and system | |
| CN115561133B (en) | Automatic identification method and system for abnormal data during CEMS calibration in thermal power industry |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |