CN114017136A - Soft light character plate alarm method and system for combined cycle generator set - Google Patents
Soft light character plate alarm method and system for combined cycle generator set Download PDFInfo
- Publication number
- CN114017136A CN114017136A CN202111290408.3A CN202111290408A CN114017136A CN 114017136 A CN114017136 A CN 114017136A CN 202111290408 A CN202111290408 A CN 202111290408A CN 114017136 A CN114017136 A CN 114017136A
- Authority
- CN
- China
- Prior art keywords
- alarm
- value
- sub
- dynamic threshold
- load
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 48
- 239000011159 matrix material Substances 0.000 claims abstract description 51
- 238000001914 filtration Methods 0.000 claims abstract description 36
- 239000013598 vector Substances 0.000 claims abstract description 32
- 238000005259 measurement Methods 0.000 claims abstract description 31
- 238000010248 power generation Methods 0.000 claims abstract description 24
- 238000012549 training Methods 0.000 claims description 27
- 238000004422 calculation algorithm Methods 0.000 claims description 18
- 238000012545 processing Methods 0.000 claims description 18
- 230000002159 abnormal effect Effects 0.000 claims description 10
- 230000008569 process Effects 0.000 description 14
- 238000007635 classification algorithm Methods 0.000 description 12
- 230000006870 function Effects 0.000 description 9
- 230000008901 benefit Effects 0.000 description 3
- 230000008859 change Effects 0.000 description 3
- 230000006872 improvement Effects 0.000 description 3
- 238000010801 machine learning Methods 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 230000003068 static effect Effects 0.000 description 3
- 238000004590 computer program Methods 0.000 description 2
- 230000003993 interaction Effects 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 230000001364 causal effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 239000000835 fiber Substances 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000005096 rolling process Methods 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
Images
Classifications
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F01—MACHINES OR ENGINES IN GENERAL; ENGINE PLANTS IN GENERAL; STEAM ENGINES
- F01D—NON-POSITIVE DISPLACEMENT MACHINES OR ENGINES, e.g. STEAM TURBINES
- F01D21/00—Shutting-down of machines or engines, e.g. in emergency; Regulating, controlling, or safety means not otherwise provided for
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F01—MACHINES OR ENGINES IN GENERAL; ENGINE PLANTS IN GENERAL; STEAM ENGINES
- F01D—NON-POSITIVE DISPLACEMENT MACHINES OR ENGINES, e.g. STEAM TURBINES
- F01D21/00—Shutting-down of machines or engines, e.g. in emergency; Regulating, controlling, or safety means not otherwise provided for
- F01D21/003—Arrangements for testing or measuring
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F02—COMBUSTION ENGINES; HOT-GAS OR COMBUSTION-PRODUCT ENGINE PLANTS
- F02C—GAS-TURBINE PLANTS; AIR INTAKES FOR JET-PROPULSION PLANTS; CONTROLLING FUEL SUPPLY IN AIR-BREATHING JET-PROPULSION PLANTS
- F02C6/00—Plural gas-turbine plants; Combinations of gas-turbine plants with other apparatus; Adaptations of gas- turbine plants for special use
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
Abstract
A soft light word board alarm method and system of a combined cycle generator set, read the historical data from the real-time database as the primitive sample set and carry on the initial filtration according to the table of measurement; performing single classification on the sub-sample sets by secondary filtering respectively, and removing outliers to form a healthy sub-sample set; respectively selecting measuring point samples which are not less than a first preset multiple in the measuring point table from the healthy sub-sample set to form sub-memory matrixes which are respectively suitable for corresponding load sections; forming a time sequence by using the corresponding single-classification support vectors for each load segment, and taking the sample standard deviation of a second preset multiple of the time sequence as a dynamic threshold; matching and selecting the sub memory matrix and the dynamic threshold according to the load section where the power generation load is located; weighting and calculating an estimated value of a given alarm parameter by taking the similarity as weight to serve as an operation normal value; and judging whether to alarm or not according to the absolute value of the deviation and the current dynamic threshold value. The invention can realize the dynamic alarm of the soft light word board to the analog quantity parameter.
Description
Technical Field
The invention relates to a soft light character plate alarm method and system of a combined cycle generator set, and belongs to the technical field of thermal parameter fault diagnosis.
Background
The combined cycle generator set mainly depends on list alarm information displayed by a Distributed Control System (DCS) in a rolling mode, is used as a primary technical measure for thermal monitoring, is particularly important for the start-up, stop processes and operation of the unit, and is also provided with a backup operation means independent of the DCS. In addition, the light word tablet that signal source and DCS are independent reports to the police, utilizes sound, light signal to remind operating personnel in time to monitor and handle fault information, is one of generating set's reserve supervisory equipment.
According to the requirements of related technical regulations, a newly-built power plant is not forced to be provided with a light word board alarm device any more, and the traditional light word board is gradually replaced by a 'soft light word board' taking DCS data as a signal source. The soft light word plate alarm needs to be logically configured by using a computer with engineer station function provided by a DCS manufacturer. Compared with the traditional light word plate, the soft light word plate has the advantage that all data of DCS can be collected. The soft light character plate can not only display alarm information on a large screen of the centralized control room according to the traditional light character plate display style and means including sound and light, but also can be configured with richer pictures in DCS.
At present, the alarm of soft light word board is used to monitor whether some important analog quantity parameters deviate from normal operation values in operation. Although the standard values (or called as the response value, the target value, the reference value and the like) of some alarm parameters in operation can be obtained by methods such as mechanism analysis, variable working condition algorithm, machine learning and the like, the purpose of the standard values is to optimize the economic performance of the unit and cannot reflect the change of the characteristics of the unit equipment in the aspects of mechanics, machinery, electricity and the like, so that the standard values are not suitable for being used as normal values for soft light word plate alarm. Therefore, the soft light word plate alarm system usually has no normal value of the alarm parameter, and the high limit value and the low limit value of the alarm parameter are commonly used for expressing the maximum deviation degree allowed in the operation. When the parameter value is between the upper limit and the lower limit, the soft light word board system is considered to be normal, and when the parameter value is higher than the upper limit or lower than the lower limit (generally called as 'overrun'), the soft light word board system gives an alarm.
At the present stage, a newly built thermal power generating unit starts to deploy a plant-level monitoring information system (SIS), and the SIS mainly has the function of gathering data and graphs of various control systems which are independently dispersed in a power plant together so as to eliminate information isolated islands and can also be developed secondarily based on the SIS. SIS manufacturers generally copy only the content and logic of the DCS soft light word board alarm into SIS completely without making initiative improvement.
The DCS or SIS-based soft light word board alarm system has the defects that the high limit and the low limit of a certain alarm parameter are static values, the distance between the high limit and the low limit is quite wide, and the static alarm mode causes the soft light word board alarm system to alarm only serious faults which are not allowed by regulations. The inadequacies of the static alarm mode have long been recognized in the power generation industry, and it is desirable to improve soft light word plate alarms in a dynamic alarm mode, i.e., running normal plus dynamic thresholds. Therefore, the dynamic alarm of the alarm parameters of the generator set by using the soft light word plate is a technical problem which needs to be solved urgently.
Disclosure of Invention
Therefore, the invention provides a soft light word plate alarm method and system of a combined cycle generator set, which realize the acquisition of the dynamic operation normal value and the dynamic threshold value of a soft light word plate alarm system and achieve the dynamic alarm of the analog quantity parameters of the soft light word plate.
In order to achieve the above purpose, the invention provides the following technical scheme: a soft light word board alarm method of a combined cycle generator set comprises off-line training and on-line estimation;
the offline training includes:
acquiring a measuring point table for constructing a given alarm parameter alarm model, reading historical data from a power plant real-time database according to the measuring point table, taking the historical data as an original sample set, and initially filtering abnormal samples in the original sample set;
dividing the original sample set into a plurality of sub-sample sets of load sections according to the capacity of the unit and the power generation load; performing secondary filtration on the sub-sample sets, performing single classification on the sub-sample sets of each load section in the secondary filtration process, and removing outliers to form a plurality of healthy sub-sample sets;
respectively selecting measuring point samples which are not less than a first preset multiple in the measuring point table from the healthy subsample set of the plurality of load sections to form a submaster memory matrix which is respectively suitable for the corresponding load sections;
forming a time sequence by using the corresponding single classification support vectors for each load segment, and taking the sample standard deviation of a second preset multiple of the time sequence as a dynamic threshold;
the online estimation comprises:
reading real-time measurement values in the measurement point table of the alarm model on line, and taking the real-time measurement values as new samples; matching and selecting the sub memory matrix and the dynamic threshold according to the load section where the power generation load is located;
calculating the similarity of the new sample and all samples in the sub memory matrix, weighting and calculating the estimated value of the given alarm parameter by taking the similarity as weight, and taking the estimated value as the normal running value of the given alarm parameter;
and calculating the deviation between the real-time measured value and the estimated value of the given alarm parameter, and judging whether to alarm or not according to the absolute value of the deviation and the current dynamic threshold value.
As a preferred scheme of a soft light word board alarm method of a combined cycle generator set, reading historical data of a preset time period from a power plant real-time database according to a measuring point table, and transferring the historical data into a CSV format as an original sample set;
and filtering abnormal samples in the original sample set by taking the high limit and the low limit of the given alarm parameters specified by the operating regulations as initial filtering conditions.
As a preferred scheme of the soft light word board alarm method of the combined cycle generator set, the sub-memory matrix comprises the maximum value and the minimum value of each measuring point, and the measured values of each measuring point except the maximum value and the minimum value are randomly selected.
As a preferred scheme of the soft light word board alarm method of the combined cycle generator set, the similarity between the new sample and all samples in the sub memory matrix is calculated through a multivariate state estimation algorithm, and the similarity is used as a weight to calculate the estimated value of the given alarm parameter in a weighting mode.
As a preferred scheme of the soft light word board alarm method of the combined cycle generator set, when the absolute value of the deviation exceeds the current dynamic threshold, the state is judged to be in a non-healthy state, and an alarm is given; and when the absolute value of the deviation does not exceed the current dynamic threshold, judging the state as a healthy state, and not giving an alarm.
As a preferred scheme of the soft light word plate alarm method of the combined cycle generator set, each of the high load section, the medium load section and the low load section has a current time sequence of the given alarm parameter real-time value, and each time sequence has a current dynamic threshold value;
when a given alarm parameter meets an alarm condition, keeping the time sequence of the real-time value of the given alarm parameter and the dynamic threshold unchanged;
updating the time series of the real-time values of the given alarm parameter and updating the dynamic threshold when the given alarm parameter does not satisfy an alarm condition.
The invention also provides a soft light word board alarm system of the combined cycle generator set, which comprises an off-line training unit and an on-line estimation unit;
the offline training unit includes:
the system comprises an original sample set acquisition module, a data processing module and a data processing module, wherein the original sample set acquisition module is used for acquiring a measuring point table for constructing a given alarm parameter alarm model, reading historical data from a real-time database of a power plant according to the measuring point table and taking the historical data as an original sample set;
the exception handling module is used for carrying out initial filtering on the exception samples in the original sample set;
the sub-sample set processing module is used for dividing the original sample set into a plurality of sub-sample sets of load sections according to the unit capacity and the power generation load; performing secondary filtration on the sub-sample sets, performing single classification on the sub-sample sets of each load section in the secondary filtration process, and removing outliers to form a plurality of healthy sub-sample sets;
the sub memory matrix processing module is used for respectively selecting the measuring point samples which are not less than a first preset multiple in the measuring point table from the sub sample sets of the plurality of load sections to form sub memory matrices which are respectively suitable for the corresponding load sections;
the dynamic threshold acquisition module is used for forming a time sequence by using the corresponding single classification support vectors for each load segment and taking the sample standard deviation of a second preset multiple of the time sequence as a dynamic threshold;
the online estimation unit includes:
the new sample module is used for reading real-time measurement values in the measurement point table of the alarm model on line and taking the real-time measurement values as new samples;
the matching selection module is used for matching and selecting the sub memory matrix and the dynamic threshold according to the load section where the power generation load is located;
a normal value obtaining module, configured to calculate similarities between the new sample and all samples in the sub-memory matrix, weight and calculate an estimated value of the given alarm parameter by using the similarities as weights, and use the estimated value as an operating normal value of the given alarm parameter;
and the alarm analysis module is used for calculating the deviation between the real-time measured value and the estimated value of the given alarm parameter and judging whether to alarm or not according to the absolute value of the deviation and the current dynamic threshold value.
As a preferred scheme of a soft light word board alarm system of a combined cycle generator set, in the original sample set acquisition module, reading historical data of a preset time period from a power plant real-time database according to the measuring point table, and transferring the historical data into a CSV format as an original sample set;
in the exception handling module, the high limit and the low limit of a given alarm parameter specified by an operating rule are used as initial filtering conditions, and abnormal samples in the original sample set are filtered;
in the sub memory matrix processing module, the sub memory matrix comprises the maximum value and the minimum value of each measuring point, and the measured values of each measuring point except the maximum value and the minimum value are randomly selected.
As a preferred scheme of a soft light word board alarm system of a combined cycle generator set, in the normal value acquisition module, calculating the similarity between the new sample and all samples in the sub memory matrix through a multivariate state estimation algorithm, and weighting and calculating the estimated value of the given alarm parameter by taking the similarity as weight;
in the alarm analysis module, when the absolute value of the deviation exceeds the current dynamic threshold, the state is judged to be in an unhealthy state, and an alarm is given; and when the absolute value of the deviation does not exceed the current dynamic threshold, judging the state as a healthy state, and not giving an alarm.
As a preferred embodiment of the soft light word plate alarm system of the combined cycle generator set, each of the high load section, the medium load section and the low load section has a current time sequence of the real-time value of the given alarm parameter, and each of the time sequences has a current dynamic threshold;
the updating processing module is used for keeping the time sequence of the real-time value of the given alarm parameter and the dynamic threshold value unchanged when the given alarm parameter meets an alarm condition; updating the time series of the real-time values of the given alarm parameter and updating the dynamic threshold when the given alarm parameter does not satisfy an alarm condition.
The invention has the following advantages: in the off-line training process, a measuring point table for constructing a given alarm parameter alarm model is obtained, historical data are read from a power plant real-time database according to the measuring point table, and the historical data are used as an original sample set; performing initial filtering on abnormal samples in the original sample set; dividing the original sample set into a plurality of sub-sample sets of load sections according to the capacity of the unit and the power generation load; performing secondary filtration on the sub-sample sets, performing single classification on the sub-sample sets of each load section in the secondary filtration process, and removing outliers to form a plurality of healthy sub-sample sets; respectively selecting measuring point samples which are not less than a first preset multiple in a measuring point table from three healthy sub-sample sets of a high load section, a medium load section and a low load section to form sub-memory matrixes which are respectively suitable for the corresponding load sections; forming a time sequence by using the corresponding single-classification support vectors for each load segment, and taking the sample standard deviation of a second preset multiple of the time sequence as a dynamic threshold; in the online estimation process, reading a real-time measurement value in a measurement point table of the alarm model on line, and taking the real-time measurement value as a new sample; matching and selecting the sub memory matrix and the dynamic threshold according to the load section where the power generation load is located; calculating the similarity of the new sample and all samples in the sub-memory matrix, weighting and calculating the estimated value of the given alarm parameter by taking the similarity as weight, and taking the estimated value as the normal running value of the given alarm parameter; and calculating the deviation between the real-time measured value and the estimated value of the given alarm parameter, and judging whether to alarm or not according to the absolute value of the deviation and the current dynamic threshold value. The invention can provide the dynamic running normal value and the dynamic threshold value of the alarm parameter of the gas-steam combined cycle unit, thereby realizing the dynamic alarm of the soft light word board on the analog quantity parameter and having better auxiliary function on the running supervision work of the gas-steam combined cycle unit.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below. It should be apparent that the drawings in the following description are merely exemplary, and that other embodiments can be derived from the drawings provided by those of ordinary skill in the art without inventive effort.
The structures, ratios, sizes, and the like shown in the present specification are only used for matching with the contents disclosed in the specification, so that those skilled in the art can understand and read the present invention, and do not limit the conditions for implementing the present invention, so that the present invention has no technical significance, and any structural modifications, changes in the ratio relationship, or adjustments of the sizes, without affecting the functions and purposes of the present invention, should still fall within the scope of the present invention.
FIG. 1 is a schematic flow chart of a soft light alarm method for a combined cycle generator set according to an embodiment of the present invention;
fig. 2 is a schematic diagram of a soft light word plate alarm system of a combined cycle generator set provided in an embodiment of the invention.
Detailed Description
The present invention is described in terms of particular embodiments, other advantages and features of the invention will become apparent to those skilled in the art from the following disclosure, and it is to be understood that the described embodiments are merely exemplary of the invention and that it is not intended to limit the invention to the particular embodiments disclosed. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Example 1
Referring to fig. 1, a soft light word plate alarm method of a combined cycle generator set is provided, which comprises off-line training and on-line estimation;
the offline training includes:
s11, obtaining a measuring point table for constructing an alarm model with given alarm parameters, reading historical data from a power plant real-time database according to the measuring point table, taking the historical data as an original sample set, and initially filtering abnormal samples in the original sample set;
s12, dividing the original sample set into three sub-sample sets of high load, medium load and low load according to the capacity of the unit and the power generation load; carrying out secondary filtration on the high-load, medium-load and low-load sub-sample sets, respectively carrying out single classification on the high-load, medium-load and low-load sub-sample sets in the secondary filtration process, and removing outliers to form three healthy sub-sample sets of a high-load section, a medium-load section and a low-load section;
s13, selecting the measuring point samples which are not less than the first preset multiple in the measuring point table from the three healthy sub-sample sets of the high load section, the medium load section and the low load section respectively to form three sub-memory matrixes which are suitable for the high load section, the medium load section and the low load section respectively;
s14, forming a time sequence by using corresponding single classification support vectors for a high load section, a medium load section and a low load section respectively, and taking the sample standard deviation of a second preset multiple of the time sequence as a dynamic threshold;
the online estimation comprises:
s21, reading real-time measurement values in the measurement point table of the alarm model on line, and taking the real-time measurement values as new samples; matching and selecting the sub memory matrix and the dynamic threshold according to the load section where the power generation load is located;
s22, calculating the similarity of the new sample and all samples in the sub memory matrix, weighting and calculating the estimated value of the given alarm parameter by taking the similarity as weight, and taking the estimated value as the normal running value of the given alarm parameter;
s23, calculating the deviation between the real-time measured value and the estimated value of the given alarm parameter, and judging whether to alarm or not according to the absolute value of the deviation and the current dynamic threshold value.
In this embodiment, the purpose of the off-line training is, on one hand, to obtain a sub-memory matrix required by the multivariate state estimation algorithm to estimate the running normal value; another aspect is to initialize the dynamic threshold with statistics of a time series of a batch of health sample data that is most recent in time given the alarm parameter.
The online estimation has the function that for a new sample obtained online, on one hand, the operation normal value of a given alarm parameter in the new sample is estimated by using a multivariate state estimation algorithm; on the other hand, the newly estimated running normal value is compared with the stored dynamic threshold value to decide whether to alarm or not.
In the embodiment, historical data of a preset time period is read from a power plant real-time database according to the measuring point table, and the historical data is stored into a CSV format and serves as the original sample set;
and filtering abnormal samples in the original sample set by taking the high limit and the low limit of the given alarm parameters specified by the operating regulations as initial filtering conditions.
In this embodiment, in the off-line training process step S11, based on the mechanism analysis and the operation rule, the measurement point table of the alarm model with the given alarm parameter is obtained, and a sample of the historical data is obtained as an original sample set.
Each alarm model is provided with a measuring point table, and except for the alarm parameters, the parameters are parameters which are causal with the alarm parameters or parameters belonging to a subsystem or sub-equipment. The power generation load has direct influence on the fluctuation of most alarm parameters, so that the power generation load is always in a measuring point table of any alarm model. The power generation load can have one to three measuring points with different physical meanings for different types of gas-steam combined cycle units, such as one-driving-one single shaft, one-driving-one split shaft and two-driving-one unit, and a professional in a power plant can judge which power generation load should be used by a certain alarm parameter.
Specifically, according to the obtained alarm model measuring point table, historical data about 1 year is collected from an SIS database of the combined cycle generator set by using a tool or an interface provided by an SIS manufacturer, and then the historical data is transferred into data in a CSV format and called an original sample set.
Specifically, the tools and interfaces provided by the SIS manufacturer can perform initial filtering, so that the abnormal samples are filtered when the original sample set is obtained by taking the high and low limits of the given alarm parameters specified by the operating regulations as filtering conditions.
In this embodiment, in the step S12 of the off-line training process, since most of the alarm parameters change with the change of the power generation load, the original sample set is divided into a plurality of sub-samples according to the power generation load, and a model is established separately for each sub-sample, which is superior to modeling with one total sample. The method is generally divided by the percentage of the generating load relative to the designed generating capacity of the unit, wherein more than 90 percent of the generating load is a high-load section, 70 percent to 90 percent of the generating load is a medium-load section, and less than 70 percent of the generating load is a low-load section, and specific numerical values can be flexibly divided and are not limited to a high section, a medium section and a low section.
Specifically, a single classification algorithm is used for classifying each subsample set and removing outliers. Common classification algorithms are generally two-class classification, one class called positive and the other negative, with the training purpose to classify the samples. The single classification algorithm has only one class, and the training aims to eliminate outliers in the class. The sample points in the training set can be regarded as points or vectors in a certain linear space, the geometric meaning of the single classification algorithm is that the points belonging to the class form a sphere in the linear space, and all the sample points with the space distance from the center of the sphere larger than the radius of the sphere are judged as outliers. The single classification algorithm is a special support vector machine classification algorithm, and sample points distributed on a space sphere are also called support vectors and have sparsity. It is possible to determine whether new samples outside the training set are outliers using only the support vectors. The single classification algorithm is realized by an OnClassSvm method in a support vector packet svm of a machine learning library sklern of python language, a kernel function uses a default Gaussian radial basis kernel function, and other parameters also use default values.
And deleting outliers of the sub-sample sets of each original sample set by using the single classification algorithm to obtain three healthy sample subsets respectively corresponding to a high load section, a medium load section and a low load section.
In this embodiment, in the step S13 of the off-line training process, specifically, from each subset of healthy samples, samples not less than 4 times the number of the measurement points in the measurement point table of the alarm model are automatically selected, so as to form three sub-memory matrices respectively suitable for high-load, medium-load, and low-load sections.
Specifically, the sub-memory matrix is a necessary prerequisite for the multivariate state estimation algorithm. And automatically selecting the sub memory matrix in an off-line mode for the MSET algorithm on line. Machine learning generally requires that the number of training samples is 2 times greater than the number of test points, wherein 4 times is the multiple relation between the number of samples and the number of test points in modeling, and the integer is taken for programming simplicity, and more multiples such as 8 times are also feasible.
Specifically, the principle of selecting the sub-memory matrix is that the maximum value and the minimum value of each measuring point are selected inevitably, and the other measuring points are selected randomly. Due to the fact that the single classification algorithm is used, the method is firstly randomly selected from the support vector samples of the single classification algorithm.
In this embodiment, in the off-line training process step S14, the on-line estimation operation normal value needs a dynamic threshold to determine whether the operation normal value exceeds the limit. Calculating the dynamic threshold belongs to the content of an off-line training part, but the dynamic threshold needs to be initialized at the early stage of online use.
Specifically, for each load segment, only the early warning real-time values of each support vector are taken from the support vectors of the single classification algorithm to form a time sequence. Taking 4 times of the standard deviation of the time series samples as a dynamic threshold. Setting the time sequence as { y1,…,yj,…yNN is the number of samples in the above time series, and is also the number of single classification support vectors.
In particular, the sample mean
Comprises the following steps:
sample variance s2Comprises the following steps:
the sample standard deviation is positive s. According to statistical theory, for normally distributed samples, 99.73% of the total number of samples in the sample battle within ± 3 times of the standard deviation of the mean. In the embodiment, 4 times of standard deviation is selected, so that the false alarm rate in online operation can be reduced.
In this embodiment, in the online estimation process, the alarm model after offline training is deployed on a server networked with the SIS. And reading real-time data of the alarm model measuring point table by using an interface provided by an SIS manufacturer on line to form a new sample vector. And automatically selecting an applicable sub memory matrix and a dynamic threshold according to the real-time value of the load and the load segmentation point.
In this embodiment, in the online estimation step S22, the similarity between the new sample and all samples in the sub memory matrix is calculated by using a multivariate state estimation algorithm, and the estimated value of the given alarm parameter is calculated by weighting with the similarity as a weight.
Specifically, the multivariate state estimation algorithm is used for calculating the new sample and the sub-memory matrix to obtain an estimated value of a given alarm parameter in the new sample, and the estimated value is used as an operation normal value. Because no special packet for realizing the MSET algorithm exists in python, the MSET algorithm is realized through programming after the memory matrix is obtained.
Specifically, the mathematical description of the MSET algorithm is as follows:
setting an alarm model with m measuring points including alarm parameters, and setting a memory matrix with n samples, wherein n is 4m,
the m measurement points observed at a history time j are called a sample j and are written in a column vector form:
wherein, the vector element xi(j) Is a scalar, i is 1,2 …, m, the column of the sub memory matrix D is composed of n history samples x (j), the matrix form of D is:
recording of the New sample
Construction of an estimate of a new sample by a sub-memory matrix D
Wherein the optimal weight vector
By minimizing residual error
To evaluate, one can obtain:
wherein D isτD reflects the dot-product relationship between two historical observation vectors in the process memory matrix, and
the point-product relationship between the new input observation vector and the historical observation vector in the process memory matrix is reflected.
The above is the traditional linear state estimation algorithm, and uses the nonlinear operator instead
The nonlinear state estimation algorithm is used for replacing dot product operation and meets several basic properties.
Representing an arbitrary non-linear operator.
The technical scheme uses Euclidean distance as
An operator. Euclidean distance is the most common nonlinear operator just satisfying nonlinear state estimation algorithm pair
All requirements of operator, therefore
dot-by-D inτD and
are changed to euclidean distances.
Two n-dimensional column vectors
The Euclidean distance of (c) is:
weight vector
Becomes the following formula:
estimating a vector
The method comprises the following steps:
euclidean distance is a measure of similarity between two vectors, the calculated weight vector
The similarity degree between the new sample and each historical sample in the sub memory matrix is shown, so the MSET method is called the similarity principle, and the higher the similarity degree between the sample of the sub memory matrix and the new sample is, the larger the weight is. Thereby estimating the vector
The value of the medium alarm parameter component is the dynamic normal operation value required by the alarm system.
In this embodiment, in the online estimation step S23, when the absolute value of the deviation exceeds the current dynamic threshold, it is determined as an unhealthy state, and an alarm is issued; and when the absolute value of the deviation does not exceed the current dynamic threshold, judging the state as a healthy state, and not giving an alarm.
In this embodiment, the online estimation step further includes S24, when a given alarm parameter satisfies an alarm condition, keeping the time series of the real-time values of the given alarm parameter and the dynamic threshold unchanged;
updating the time series of the real-time values of the given alarm parameter and updating the dynamic threshold when the given alarm parameter does not satisfy an alarm condition.
Specifically, each of the high load segment, the medium load segment and the low load segment has a current said time series of real-time values of said given alarm parameter, and each of said time series has a current said dynamic threshold. When the alarm parameters meet the alarm conditions, the new samples are not in a healthy working condition, and the real-time value time sequence and the dynamic threshold of the alarm parameters of the load section are kept unchanged. When the alarm parameter does not meet the alarm condition, the new sample is a healthy condition, the real-time value time sequence of the alarm parameter should be updated first, and then the dynamic threshold value should be updated.
Specifically, the alarm real-time value time sequence is composed of alarm parameter real-time values of support vectors of a single classification algorithm at the beginning, and each real-time value has a time tag. The new samples which do not alarm are added into the alarm real-time value time sequence one by one, the earliest support vector of the time label is correspondingly deleted from the time sequence according to the principle of first-in first-out, and the number of the samples of the time sequence is kept unchanged. So that the alarm real-time value time series consists of the latest alarm parameter real-time values very quickly.
Specifically, the method for updating the dynamic threshold is the same as the details of step S14 in the off-line training process, and 4 times of the standard deviation of the updated early warning real-time value time series is used as the next dynamic threshold.
In conclusion, in the off-line training process, a measuring point table for constructing an alarm model with given alarm parameters is obtained, historical data is read from a real-time database of a power plant according to the measuring point table, and the historical data is used as an original sample set; dividing an original sample set into three sub-sample sets of high load, medium load and low load according to the capacity of a unit and the power generation load; carrying out secondary filtration on the high-load, medium-load and low-load sub-sample sets, respectively carrying out single classification on the high-load, medium-load and low-load sub-sample sets in the secondary filtration process, and removing outliers to form three healthy sub-sample sets of a high-load section, a medium-load section and a low-load section; respectively selecting measuring point samples which are not less than a first preset multiple in a measuring point table from three healthy sub-sample sets of a high load section, a medium load section and a low load section to form three sub-memory matrixes which are respectively suitable for the high load section, the medium load section and the low load section; respectively forming a time sequence for the high load section, the medium load section and the low load section by using corresponding single-classification support vectors, and taking the sample standard deviation of a second preset multiple of the time sequence as a dynamic threshold; in the online estimation process, reading a real-time measurement value in a measurement point table of the alarm model on line, and taking the real-time measurement value as a new sample; matching and selecting the sub memory matrix and the dynamic threshold according to the load section where the power generation load is located; calculating the similarity of the new sample and all samples in the sub-memory matrix, weighting and calculating the estimated value of the given alarm parameter by taking the similarity as weight, and taking the estimated value as the normal running value of the given alarm parameter; and calculating the deviation between the real-time measured value and the estimated value of the given alarm parameter, and judging whether to alarm or not according to the absolute value of the deviation and the current dynamic threshold value. The invention can provide the dynamic running normal value and the dynamic threshold value of the alarm parameter of the gas-steam combined cycle unit, thereby realizing the dynamic alarm of the soft light word board on the analog quantity parameter and having better auxiliary function on the running supervision work of the gas-steam combined cycle unit.
Example 2
Referring to fig. 2, embodiment 2 of the present invention further provides a soft light word plate alarm system of a combined cycle generator set, including an offline training unit 1 and an online estimation unit 2;
the offline training unit 1 includes:
the system comprises an original sample set acquisition module 11, a data processing module and a data processing module, wherein the original sample set acquisition module is used for acquiring a measuring point table for constructing a given alarm parameter alarm model, reading historical data from a real-time database of a power plant according to the measuring point table and taking the historical data as an original sample set;
an exception handling module 12, configured to perform initial filtering on the exception samples in the original sample set;
a sub-sample set processing module 13, configured to divide the original sample set into a sub-sample set of a plurality of load segments according to the unit capacity and the power generation load; performing secondary filtration on the sub-sample sets, performing single classification on the sub-sample sets of each load section in the secondary filtration process, and removing outliers to form a plurality of healthy sub-sample sets;
the sub memory matrix processing module 14 is used for respectively selecting the measuring point samples which are not less than the first preset multiple in the measuring point table from the three healthy sub sample sets of the high load section, the medium load section and the low load section to form three sub memory matrices which are respectively suitable for the high load section, the medium load section and the low load section;
a dynamic threshold obtaining module 15, configured to form a time sequence for the high load segment, the medium load segment, and the low load segment by using the corresponding single-class support vectors, and take a sample standard deviation of a second preset multiple of the time sequence as a dynamic threshold;
the online estimation unit 2 includes:
a new sample module 21, configured to read a real-time measurement value in the measurement point table of the alarm model on line, and use the real-time measurement value as a new sample;
the matching selection module 22 is used for matching and selecting the sub memory matrix and the dynamic threshold according to the load section where the power generation load is located;
a normal value obtaining module 23, configured to calculate similarities between the new sample and all samples in the sub-memory matrix, calculate an estimated value of the given alarm parameter by using the similarities as weights, and use the estimated value as an operating normal value of the given alarm parameter;
and the alarm analysis module 24 is configured to calculate a deviation between the real-time measurement value and the estimated value of the given alarm parameter, and determine whether to alarm according to an absolute value of the deviation and the current dynamic threshold.
In this embodiment, in the original sample set obtaining module 11, reading historical data of a preset time period from a power plant real-time database according to the measurement point table, and storing the historical data in a CSV format as the original sample set;
in the exception handling module 12, the high limit and the low limit of a given alarm parameter specified by an operation rule are used as an initial filtering condition to filter out the exception samples in the original sample set;
in the sub memory matrix processing module 14, the sub memory matrix includes the maximum value and the minimum value of each measuring point, and the measured values of each measuring point except the maximum value and the minimum value are randomly selected.
In this embodiment, in the normal value obtaining module 23, the similarity between the new sample and all samples in the sub-memory matrix is calculated through a multivariate state estimation algorithm, and the estimated value of the given alarm parameter is calculated by weighting with the similarity as a weight;
in the alarm analysis module 24, when the absolute value of the deviation exceeds the current dynamic threshold, it is determined as an unhealthy state, and an alarm is given; and when the absolute value of the deviation does not exceed the current dynamic threshold, judging the state as a healthy state, and not giving an alarm.
In this embodiment, each of the high load segment, the medium load segment and the low load segment has a current time sequence of the real-time value of the given alarm parameter, and each of the time sequences has a current dynamic threshold;
the system further comprises an update processing module 25, configured to keep the time series of the real-time values of the given alarm parameter and the dynamic threshold unchanged when the given alarm parameter satisfies an alarm condition; updating the time series of the real-time values of the given alarm parameter and updating the dynamic threshold when the given alarm parameter does not satisfy an alarm condition.
It should be noted that, for the information interaction, execution process, and other contents between the modules/units of the system, since the same concept is based on the method embodiment in embodiment 1 of the present application, the technical effect brought by the information interaction, execution process, and other contents are the same as those of the method embodiment of the present application, and specific contents may refer to the description in the foregoing method embodiment of the present application, and are not described herein again.
Example 3
Embodiment 3 of the present invention provides a non-transitory computer-readable storage medium having stored therein program code for a soft light word plate alarm method of a combined cycle generator set, the program code including instructions for performing the soft light word plate alarm method of the combined cycle generator set of embodiment 1 or any possible implementation thereof.
The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device, such as a server, a data center, etc., that incorporates one or more of the available media. The usable medium may be a magnetic medium (e.g., floppy Disk, hard Disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., Solid State Disk (SSD)), among others.
Example 4
An embodiment 4 of the present invention provides an electronic device, including: a memory and a processor;
the processor and the memory are communicated with each other through a bus; the memory stores program instructions executable by the processor, the processor invoking the program instructions to perform the soft light word plate alarm method of the combined cycle generator set of embodiment 1 or any possible implementation thereof.
Specifically, the processor may be implemented by hardware or software, and when implemented by hardware, the processor may be a logic circuit, an integrated circuit, or the like; when implemented in software, the processor may be a general-purpose processor implemented by reading software code stored in a memory, which may be integrated in the processor, located external to the processor, or stand-alone.
In the above embodiments, the implementation may be wholly or partially realized by software, hardware, firmware, or any combination thereof. When implemented in software, may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When loaded and executed on a computer, cause the processes or functions described in accordance with the embodiments of the invention to occur, in whole or in part. The computer may be a general purpose computer, a special purpose computer, a network of computers, or other programmable device. The computer instructions may be stored in a computer readable storage medium or transmitted from one computer readable storage medium to another, for example, from one website site, computer, server, or data center to another website site, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, Digital Subscriber Line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.).
It will be apparent to those skilled in the art that the modules or steps of the present invention described above may be implemented by a general purpose computing device, they may be centralized on a single computing device or distributed across a network of multiple computing devices, and alternatively, they may be implemented by program code executable by a computing device, such that they may be stored in a storage device and executed by a computing device, and in some cases, the steps shown or described may be performed in an order different than that described herein, or they may be separately fabricated into individual integrated circuit modules, or multiple ones of them may be fabricated into a single integrated circuit module. Thus, the present invention is not limited to any specific combination of hardware and software.
Although the invention has been described in detail above with reference to a general description and specific examples, it will be apparent to one skilled in the art that modifications or improvements may be made thereto based on the invention. Accordingly, such modifications and improvements are intended to be within the scope of the invention as claimed.
Claims (10)
1. A soft light word board alarm method of a combined cycle generator set is characterized by comprising off-line training and on-line estimation;
the offline training includes:
acquiring a measuring point table for constructing a given alarm parameter alarm model, reading historical data from a power plant real-time database according to the measuring point table, taking the historical data as an original sample set, and initially filtering abnormal samples in the original sample set;
dividing the original sample set into a plurality of sub-sample sets of load sections according to the capacity of the unit and the power generation load; performing secondary filtration on the sub-sample sets, performing single classification on the sub-sample sets of each load section in the secondary filtration process, and removing outliers to form a plurality of healthy sub-sample sets;
respectively selecting measuring point samples which are not less than a first preset multiple in the measuring point table from the healthy subsample set of the plurality of load sections to form a submaster memory matrix which is respectively suitable for the corresponding load sections;
forming a time sequence by using the corresponding single classification support vectors for each load segment, and taking the sample standard deviation of a second preset multiple of the time sequence as a dynamic threshold;
the online estimation comprises:
reading real-time measurement values in the measurement point table of the alarm model on line, and taking the real-time measurement values as new samples; matching and selecting the sub memory matrix and the dynamic threshold according to the load section where the power generation load is located;
calculating the similarity of the new sample and all samples in the sub memory matrix, weighting and calculating the estimated value of the given alarm parameter by taking the similarity as weight, and taking the estimated value as the normal running value of the given alarm parameter;
and calculating the deviation between the real-time measured value and the estimated value of the given alarm parameter, and judging whether to alarm or not according to the absolute value of the deviation and the current dynamic threshold value.
2. The soft light word board alarm method of the combined cycle generator set according to claim 1, wherein historical data of a preset time period is read from a power plant real-time database according to the measuring point table, and the historical data is saved into a CSV format as the original sample set;
and filtering abnormal samples in the original sample set by taking the high limit and the low limit of the given alarm parameters specified by the operating regulations as initial filtering conditions.
3. The soft light word plate alarm method of the combined cycle generator set as claimed in claim 1, wherein the sub memory matrix comprises a maximum value and a minimum value of each measuring point, and the measured values of each measuring point except the maximum value and the minimum value are randomly selected.
4. The soft light word plate alarm method of the combined cycle generator set according to claim 1, wherein the similarity between the new sample and all samples in the sub memory matrix is calculated through a multivariate state estimation algorithm, and the estimated value of the given alarm parameter is calculated by weighting with the similarity as a weight.
5. The soft light word plate alarm method of the combined cycle generator set according to claim 1, characterized in that when the absolute value of the deviation exceeds the current dynamic threshold, the state is judged as unhealthy, and an alarm is given; and when the absolute value of the deviation does not exceed the current dynamic threshold, judging the state as a healthy state, and not giving an alarm.
6. A soft light word plate alarm method in a combined cycle generator set according to claim 5, wherein each load segment has a current said time series of real time values of said given alarm parameter, each said time series having a current said dynamic threshold;
when a given alarm parameter meets an alarm condition, keeping the time sequence of the real-time value of the given alarm parameter and the dynamic threshold unchanged;
updating the time series of the real-time values of the given alarm parameter and updating the dynamic threshold when the given alarm parameter does not satisfy an alarm condition.
7. A soft light word board alarm system of a combined cycle generator set is characterized by comprising an off-line training unit and an on-line estimation unit;
the offline training unit includes:
the system comprises an original sample set acquisition module, a data processing module and a data processing module, wherein the original sample set acquisition module is used for acquiring a measuring point table for constructing a given alarm parameter alarm model, reading historical data from a real-time database of a power plant according to the measuring point table and taking the historical data as an original sample set;
the exception handling module is used for carrying out initial filtering on the exception samples in the original sample set;
the sub-sample set processing module is used for dividing the original sample set into a plurality of sub-sample sets of load sections according to the unit capacity and the power generation load; performing secondary filtration on the sub-sample sets, performing single classification on the sub-sample sets of each load section in the secondary filtration process, and removing outliers to form a plurality of healthy sub-sample sets;
the sub memory matrix processing module is used for respectively selecting the measuring point samples which are not less than a first preset multiple in the measuring point table from the sub sample sets of the plurality of load sections to form sub memory matrices which are respectively suitable for the corresponding load sections;
the dynamic threshold acquisition module is used for forming a time sequence by using the corresponding single classification support vectors for each load segment and taking the sample standard deviation of a second preset multiple of the time sequence as a dynamic threshold;
the online estimation unit includes:
the new sample module is used for reading real-time measurement values in the measurement point table of the alarm model on line and taking the real-time measurement values as new samples;
the matching selection module is used for matching and selecting the sub memory matrix and the dynamic threshold according to the load section where the power generation load is located;
a normal value obtaining module, configured to calculate similarities between the new sample and all samples in the sub-memory matrix, weight and calculate an estimated value of the given alarm parameter by using the similarities as weights, and use the estimated value as an operating normal value of the given alarm parameter;
and the alarm analysis module is used for calculating the deviation between the real-time measured value and the estimated value of the given alarm parameter and judging whether to alarm or not according to the absolute value of the deviation and the current dynamic threshold value.
8. The soft light word board alarm system of the combined cycle generator set according to claim 7, wherein the original sample set acquisition module reads historical data of a preset time period from a power plant real-time database according to the measuring point table, and the historical data is saved into a CSV format as the original sample set;
in the exception handling module, the high limit and the low limit of a given alarm parameter specified by an operating rule are used as initial filtering conditions, and abnormal samples in the original sample set are filtered;
in the sub memory matrix processing module, the sub memory matrix comprises the maximum value and the minimum value of each measuring point, and the measured values of each measuring point except the maximum value and the minimum value are randomly selected.
9. The soft light word plate alarm system of a combined cycle generator set of claim 7, wherein in the normal value obtaining module, the similarity between the new sample and all samples in the sub memory matrix is calculated through a multivariate state estimation algorithm, and the similarity is used as a weight to calculate the estimated value of the given alarm parameter;
in the alarm analysis module, when the absolute value of the deviation exceeds the current dynamic threshold, the state is judged to be in an unhealthy state, and an alarm is given; and when the absolute value of the deviation does not exceed the current dynamic threshold, judging the state as a healthy state, and not giving an alarm.
10. A soft light word plate alarm system for a combined cycle generator set as claimed in claim 7, wherein each load segment has a current said time series of real time values of said given alarm parameter, each said time series having a current said dynamic threshold;
the updating processing module is used for keeping the time sequence of the real-time value of the given alarm parameter and the dynamic threshold value unchanged when the given alarm parameter meets an alarm condition; updating the time series of the real-time values of the given alarm parameter and updating the dynamic threshold when the given alarm parameter does not satisfy an alarm condition.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111290408.3A CN114017136A (en) | 2021-11-02 | 2021-11-02 | Soft light character plate alarm method and system for combined cycle generator set |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111290408.3A CN114017136A (en) | 2021-11-02 | 2021-11-02 | Soft light character plate alarm method and system for combined cycle generator set |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114017136A true CN114017136A (en) | 2022-02-08 |
Family
ID=80059801
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111290408.3A Pending CN114017136A (en) | 2021-11-02 | 2021-11-02 | Soft light character plate alarm method and system for combined cycle generator set |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114017136A (en) |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150051847A1 (en) * | 2011-11-22 | 2015-02-19 | Electric Power Research Institute, Inc. | System and method for anomaly detection |
| CN105114348A (en) * | 2015-09-09 | 2015-12-02 | 长春合成兴业能源技术有限公司 | Device and method for induced draft fan fault early warning based on power station operation data |
| CN105302848A (en) * | 2014-10-11 | 2016-02-03 | 山东鲁能软件技术有限公司 | Evaluation value calibration method of equipment intelligent early warning system |
| CN107489464A (en) * | 2017-07-20 | 2017-12-19 | 中国神华能源股份有限公司 | Turbo-generator Sets Faults method for early warning and system |
| US20180307997A1 (en) * | 2017-04-20 | 2018-10-25 | General Electric Company | Systems and methods for improved quantification of uncertainty in turbomachinery |
| US20180313721A1 (en) * | 2015-11-05 | 2018-11-01 | Safran Aircraft Engines | System and method for monitoring a turbomachine, with indicator merging for the synthesis of an alarm confirmation |
| CN108897954A (en) * | 2018-06-29 | 2018-11-27 | 龙源（北京）风电工程技术有限公司 | Wind turbines temperature pre-warning method and its system based on BootStrap confidence calculations |
| CN112530526A (en) * | 2020-12-09 | 2021-03-19 | 浙江浙能北仑发电有限公司 | Fly ash carbon content prediction method and device and electronic equipment |
-
2021
- 2021-11-02 CN CN202111290408.3A patent/CN114017136A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150051847A1 (en) * | 2011-11-22 | 2015-02-19 | Electric Power Research Institute, Inc. | System and method for anomaly detection |
| CN105302848A (en) * | 2014-10-11 | 2016-02-03 | 山东鲁能软件技术有限公司 | Evaluation value calibration method of equipment intelligent early warning system |
| CN105114348A (en) * | 2015-09-09 | 2015-12-02 | 长春合成兴业能源技术有限公司 | Device and method for induced draft fan fault early warning based on power station operation data |
| US20180313721A1 (en) * | 2015-11-05 | 2018-11-01 | Safran Aircraft Engines | System and method for monitoring a turbomachine, with indicator merging for the synthesis of an alarm confirmation |
| US20180307997A1 (en) * | 2017-04-20 | 2018-10-25 | General Electric Company | Systems and methods for improved quantification of uncertainty in turbomachinery |
| CN107489464A (en) * | 2017-07-20 | 2017-12-19 | 中国神华能源股份有限公司 | Turbo-generator Sets Faults method for early warning and system |
| CN108897954A (en) * | 2018-06-29 | 2018-11-27 | 龙源（北京）风电工程技术有限公司 | Wind turbines temperature pre-warning method and its system based on BootStrap confidence calculations |
| CN112530526A (en) * | 2020-12-09 | 2021-03-19 | 浙江浙能北仑发电有限公司 | Fly ash carbon content prediction method and device and electronic equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10809704B2 (en) | Process performance issues and alarm notification using data analytics | |
| US20210019211A1 (en) | Method and device for determining a performance indicator value for predicting anomalies in a computing infrastructure from values of performance indicators | |
| US10635096B2 (en) | Methods for analytics-driven alarm rationalization, assessment of operator response, and incident diagnosis and related systems | |
| US11657121B2 (en) | Abnormality detection device, abnormality detection method and computer readable medium | |
| JP6951101B2 (en) | Information processing device, control device, control method of information processing device, control program, and recording medium | |
| CN111459700A (en) | Method and apparatus for diagnosing device failure, diagnostic device, and storage medium | |
| JP6601433B2 (en) | Event analysis device, event analysis system, event analysis method, event analysis program, and recording medium | |
| JP6538057B2 (en) | Computer-implemented method and system for automatically monitoring and determining the status of all process segments in a process unit | |
| JP2020008997A (en) | Abnormality detection system | |
| JP7180985B2 (en) | Diagnostic device and diagnostic method | |
| CN115017214B (en) | Hydropower station auxiliary equipment operation state analysis early warning method, device and storage medium | |
| CN113988325A (en) | Power system fault early warning method and device, terminal equipment and storage medium | |
| CN109660407A (en) | Distributed system monitoring system and method | |
| KR101960755B1 (en) | Method and apparatus of generating unacquired power data | |
| CN110162743A (en) | A kind of data administering method based on k neighborhood nonlinear state Eq algorithm | |
| CN114017136A (en) | Soft light character plate alarm method and system for combined cycle generator set | |
| CN113806615A (en) | KPI (Key performance indicator) abnormity early warning method of intelligent IT operation and maintenance system | |
| CN112416896A (en) | Data abnormity warning method and device, storage medium and electronic device | |
| CN114064413A (en) | Container group adjusting and controlling method and device and electronic equipment | |
| CN115858606A (en) | Method, device and equipment for detecting abnormity of time series data and storage medium | |
| JP7452509B2 (en) | Monitoring devices, monitoring methods, and monitoring programs | |
| JP7371802B1 (en) | Abnormality diagnosis system, abnormality diagnosis device, abnormality diagnosis method, and program | |
| JP7052914B1 (en) | Abnormality diagnosis system, abnormality diagnosis device, abnormality diagnosis method, and program | |
| TWI820932B (en) | System and method for simulating equipment twin data | |
| JP2017076164A (en) | Apparatus monitoring device and alert information management method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |