CN106802643A - failure prediction system and method - Google Patents
failure prediction system and method Download PDFInfo
- Publication number
- CN106802643A CN106802643A CN201510837616.9A CN201510837616A CN106802643A CN 106802643 A CN106802643 A CN 106802643A CN 201510837616 A CN201510837616 A CN 201510837616A CN 106802643 A CN106802643 A CN 106802643A
- Authority
- CN
- China
- Prior art keywords
- failure
- prediction
- event
- power plant
- rule
- 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
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B23/00—Testing or monitoring of control systems or parts thereof
- G05B23/02—Electric testing or monitoring
- G05B23/0205—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults
- G05B23/0218—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterised by the fault detection method dealing with either existing or incipient faults
- G05B23/0224—Process history based detection method, e.g. whereby history implies the availability of large amounts of data
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B2219/00—Program-control systems
- G05B2219/20—Pc systems
- G05B2219/24—Pc safety
- G05B2219/24047—Count certain number of errors, faults before delivering alarm, stop
Abstract
A kind of failure prediction system includes that rule produces processing unit and detecting fault processing unit.Rule produces processing unit for determining multiple correlation rules according to the historical data in power plant.Detecting fault processing unit, for performing following operation:Receive the current alert sequence in power plant;Current alert sequence is compared with one or more the critical alert patterns in each correlation rule;And the corresponding prediction failure that event of failure occurs in being produced for preventing power plant during one or more critical alert pattern match of the warning mode and this in current alert sequence;Wherein, each critical alert pattern includes the two or more alarms related to the previous failures event in historical data.The present invention also provides a kind of failure prediction method.
Description
Technical field
The present invention relates to failure prediction system and method, more particularly to the failure prediction system in power plant and
Method.
Background technology
The industrial operation of the high complexity realized for example in power plant environment often refers to multiple machines and pass
The complicated coordination of connection process.Many industrial parts in such power plant environment may include and calculate dress
The sensor or other monitoring devices of combination are put so that the real time status of such part can be chased after by electronics
Track.For example, some display units in power plant environment can show respective with what is monitored in factory
The various current plant operating condition of part or procedure correlation.
The peration data in above-mentioned power plant usually can use only in the form of continuous time series.In other words,
The continuous monitoring component of sensor and provide ceaselessly data flow so that operator can observe various portions of factory
The real-time statistics of the present operating state of part.It is important to choose specific plant operation from these data
Thing.
Some known technologies can respond specific problem or thing and only lead on the basis of special if necessary
Cross experience and classify and look back the manual operations of information and analyze specific plant operation.Such technology typical case
Ground includes that hand digging mass data finds specific plant operation and/or event, filter these operations/
Event finds out the operations/events of correlation, takes out a little signal from data, and then according to giving each other
It is drawn.All so very long and complicated steps are general to be completed on the basis of special, and typically
Must be repeated when present for each problem.Therefore, still have automation and streamlined data point
Analysis and the demand of the event correlation occurred in the environment of plant.
Due to the vast number of the information of seizure in traditional monitoring system and classification and the such data of access
Restricted manner, analysis of history data may be relatively difficult.There is no mode recognize and store with it is past
The data of Action Events association, it is desired to recognize that operator may be forced a large amount of historical datas of manual sort
Information, compare consuming time, inefficiency.Therefore, industry is still present demand to classify and analyze
The comparing of power plant historical data and/or offer current data and historical data.
In addition, certain the class fault mode in traditional monitoring system only produces alarm, without operator
Specific correction is taken to operate, operator can ignore the grade alarm often.However, operator warns to the grade
The ignorance of report, may finally cause the damage of part in power plant or power plant may be caused to be forced shutdown
Deng these can all cause damage.
The content of the invention
One or more aspects of the invention are concluded now in order to basic comprehension of the invention, and wherein this is returned
It not is extensive overview of the invention to receive, and is not intended to mark some key elements of the invention, not yet
It is intended to mark its scope.Conversely, the main purpose of the conclusion is before more detailed description is presented below
Some concepts of the invention are presented with reduced form.
One aspect of the present invention, is to provide a kind of failure prediction system, and it includes:
Rule produces processing unit, for determining multiple correlation rules according to the historical data in power plant;And
Detecting fault processing unit, for performing following operation:
Receive the current alert sequence in power plant;
Current alert sequence is compared with one or more the critical alert patterns in each correlation rule
Compared with;And
Produced when warning mode in current alert sequence is with one or more critical alert pattern match
Correspondence prediction failure for preventing event of failure generation in power plant;
Wherein, each critical alert pattern includes two related to the previous failures event in historical data
Or more than two alarms.
It is preferred that in above-mentioned failure prediction system, the failure prediction system also includes:
Operation processing unit is corrected, operation is corrected for receiving prediction failure and being performed according to the prediction failure
To prevent the generation of event of failure in power plant.
It is preferred that in above-mentioned failure prediction system, for entangling of preventing event of failure in power plant from occurring
Power-off operation that positive operation is including power plant, the operation for changing power plant, software program repaiies in power plant
Multiple operation, regulation maintenance project or combinations thereof.
It is preferred that in above-mentioned failure prediction system, the previous failures event in the historical data includes different
Normal shutdown, normal shutdown, critical alert or combinations thereof.
It is preferred that in above-mentioned failure prediction system, the rule produces processing unit to be additionally operable to according to history
Multiple previous failures events in data and multiple previous alarm sequences determine multiple correlation rules.
It is preferred that in above-mentioned failure prediction system, it is following that the rule produces processing unit to be additionally operable to execution
Operate to determine the plurality of correlation rule:
One or more event of failure samples are identified from all previous failures events of historical data;
Calculate the frequency of occurrences and error prediction of all warning modes related to each event of failure sample
Rate;
Corresponding multiple grades are calculated according to multiple frequencies of occurrences and multiple misprediction rates;And
There to be calculating grade to be defined as critical alert mould more than any one warning mode of predetermined threshold
Formula;Wherein, each warning mode includes two or more continuous alarms.
It is preferred that in above-mentioned failure prediction system, the rule produces processing unit to be additionally operable in all events
Barrier event sample process determines that multiple correlation rules and the multiple correlation rules that will be determined send when finishing
To power station or detecting fault processing unit, each correlation rule includes one or more critical alert patterns
And its grade and the event of failure related to one or more critical alert patterns.
It is preferred that in above-mentioned failure prediction system, it is following that the rule produces processing unit to be additionally operable to execution
Operate to update the plurality of correlation rule:
Receive the new event of failure in power plant and corresponding alarm;
Recognized when new event of failure belongs to one or more event of failure samples related to the new event of failure
Current alert mode and Historical Alerts pattern;
Update the frequency of occurrences of all warning modes related to new event of failure;
Corresponding multiple grades are updated according to multiple frequencies of occurrences for updating and multiple misprediction rates;And
To be there is more New r4 to be defined as critical alert mould more than any one warning mode of predetermined threshold
Formula.
It is preferred that in above-mentioned failure prediction system, it is following that the rule produces processing unit to be additionally operable to execution
Operate to update the plurality of correlation rule:
Receive one or more warning modes of prediction failure and triggering prediction failure;
Increase the misprediction rate of one or more warning modes when predicting that failure is error prediction;
The occurrence rate of one or more warning modes is updated when predicting that failure is not error prediction;
Corresponding multiple grades or root are updated according to multiple frequencies of occurrences for updating and multiple misprediction rates
Corresponding multiple grades are updated according to the misprediction rate of multiple frequencies of occurrences and multiple increases;And
To be there is more New r4 to be defined as critical alert mould more than any one warning mode of predetermined threshold
Formula.
It is preferred that in above-mentioned failure prediction system, the error prediction processing unit is additionally operable to be warned currently
Corresponding prediction failure is produced when reporting one or more the critical alert pattern match of warning mode and this in sequence
And its possibility.
It is preferred that in above-mentioned failure prediction system, the error prediction processing unit is additionally operable to according to triggering
Predict that the grade of one or more warning modes of failure predicts the possibility of failure to calculate.
Another aspect of the present invention, is to provide a kind of failure prediction method, and it includes:
Historical data according to power plant determines multiple correlation rules;
Receive the current alert sequence in power plant;
Current alert sequence is compared with one or more the critical alert patterns in each correlation rule
Compared with;And
Produced when warning mode in current alert sequence is with one or more critical alert pattern match
Correspondence prediction failure for preventing event of failure generation in power plant;
Wherein, each critical alert pattern includes two related to the previous failures event in historical data
Or more than two alarms.
It is preferred that above-mentioned failure prediction method also includes:
Receive prediction failure and performed according to the prediction failure and correct operation to prevent event of failure in power plant
Generation;Correction operation include power plant power-off operation, change the operation in power plant, in power plant
The reparation operation of software program, regulation maintenance project or combinations thereof.
It is preferred that in above-mentioned failure prediction method, determining the method for the plurality of correlation rule includes:
One or more event of failure samples are identified from all previous failures events of historical data;
Calculate the frequency of occurrences and error prediction of all warning modes related to each event of failure sample
Rate;
Corresponding multiple grades are calculated according to multiple frequencies of occurrences and multiple misprediction rates;
There to be calculating grade to be defined as critical alert mould more than any one warning mode of predetermined threshold
Formula;And
Multiple correlation rules are determined when all event of failure sample process are finished;
Wherein, each warning mode includes two or more continuous alarms.
It is preferred that in above-mentioned failure prediction method, the method for updating the plurality of correlation rule includes:
Receive the new event of failure in power plant and corresponding alarm;
The identification current alert mode related to the new event of failure and Historical Alerts pattern;
The all alarm moulds related to new event of failure are updated when new event of failure belongs to event of failure sample
The frequency of occurrences of formula;
Corresponding multiple grades are updated according to multiple frequencies of occurrences for updating and multiple misprediction rates;And
To be there is more New r4 to be defined as critical alert mould more than any one warning mode of predetermined threshold
Formula.
It is preferred that in above-mentioned failure prediction method, the method for updating the plurality of correlation rule includes:
Receive one or more the alarm moulds in the current alert sequence of prediction failure and triggering prediction failure
Formula;
Increase the misprediction rate of one or more warning modes when predicting that failure is error prediction;
The occurrence rate of one or more warning modes is updated when predicting that failure is not error prediction;
Corresponding multiple grades or root are updated according to multiple frequencies of occurrences for updating and multiple misprediction rates
Corresponding multiple grades are updated according to the misprediction rate of multiple frequencies of occurrences and multiple increases;And
To be there is more New r4 to be defined as critical alert mould more than any one warning mode of predetermined threshold
Formula.
Failure prediction system and method that embodiment of the present invention is provided, can be to the historical data in power plant
It is analyzed to determine multiple correlation rules, by the current alert sequence in power plant and each correlation rule
One or more critical alert patterns be compared, and the warning mode in current alert sequence and one
The correspondence prediction failure in power plant is produced during individual or multiple critical alert pattern match.On the one hand, Ke Yiman
Sufficient industry is to analysis power plant historical data and/or the demand of the comparing for providing current data and historical data.
On the other hand, above-mentioned prediction failure is used to prevent the generation of true fault event in power plant, which reduces
Power plant is forced the number of times of shutdown/non-predetermined shutdown, namely the operational reliability that improve power plant.
Brief description of the drawings
It is described for embodiments of the present invention by with reference to accompanying drawing, the present invention may be better understood,
In the accompanying drawings:
Fig. 1 is a kind of schematic diagram in the power plant of exemplary embodiment of the invention.
Fig. 2 is a kind of flow chart of the method for the offline determination correlation rule of exemplary embodiment.
Fig. 3 is the flow chart of the method for the online updating correlation rule of the first exemplary embodiment.
Fig. 4 is second flow chart of the method for the online updating correlation rule of exemplary embodiment.
Fig. 5 is a kind of flow chart of the failure prediction method of exemplary embodiment.
Specific embodiment
Specific embodiment of the invention explained below, it should be pointed out that in these implementation methods
During specific descriptions, in order to carry out brief and concise description, this specification can not possibly be to actual implementation
All features of mode make description in detail.It is to be understood that in any one implementation method
Actual implementation process in, as during any one engineering project or design object, in order to
The objectives of developer are realized, the limitation related or commercially related in order to meet system usually can
Various specific decision-makings are made, and this also can be from a kind of implementation method to another embodiment
Change.Although moreover, it is to be understood that effort done in this development process may
It is complicated and tediously long, but the ordinary skill of this area related for present disclosure
For personnel, some designs carried out on the basis of the technology contents that the disclosure is disclosed, manufacture or raw
The change such as product is conventional technology, is not construed as content of this disclosure insufficient.
Unless otherwise defined, the technical term or scientific terminology for being used in claims and specification should
There is in by the technical field of the invention the ordinary meaning that the personage of general technical ability understands.The present invention
" first ", " second " that is used in patent application specification and claims and similar word are simultaneously
Any order, quantity or importance are not indicated that, and is used only to distinguish different parts." one "
Or the similar word such as " " is not offered as quantity limitation, but expression has at least one." including " or
The similar word such as person's "comprising" mean to appear in " including " or "comprising" before element or object contain
Lid appear in " including " or "comprising" presented hereinafter element or object and its equivalent element, do not arrange
Except other elements or object.The similar word such as " connection " or " connected " be not limited to physics or
The connection of person's machinery, and can be including electric connection, either directly still indirectly.
Fig. 1 is referred to, it is a kind of showing for the failure prediction system 900 of exemplary embodiment of the invention
It is intended to.Failure prediction system 900 is used to produce the prediction failure in power plant 800 to prevent power plant 800
The generation of middle true fault event.Failure prediction system 900 includes that rule produces processing unit 902, event
Barrier detecting processing unit 904 and correct operation processing unit 906.
In a kind of infinite embodiment, power plant 800 includes gas turbine 100 and generator 810.
In another embodiment, power plant 800 be gas turbine, steam turbine and generating set into connection
Close cycle power plant.In addition, in other embodiments, power plant 800 can also be wind power plant,
Hydraulic power plant, solar power plant or coal-fired power plant etc..
In the present embodiment, gas turbine 100 includes compressor 120, burner 140, turbine
150 and rotating shaft 160.Compressor 120 is used for compressed air stream 122, and the air stream 122 that will be compressed is defeated
Deliver to burner 140.Burner 140 is used for the The fuel stream 142 of the air stream 122 and supercharging that will be compressed
Mixed, then light said mixture to produce burning gases stream 144.Although Fig. 1 illustrate only
One burner 140, but gas turbine 100 can include any number of burner 140.
Burning gases stream 144 is then delivered to turbine 150, burning gases stream 144 drives turbine
150, to produce mechanical energy.Turbine 150 is operationally mechanical coupling to compressor by rotating shaft 160
120 and turbine 150 is mechanical coupling to generator 112 so that the machine produced in turbine 150
Tool energy drives compressor 120 and generator 810 via rotating shaft 160, so that electric energy can be by turbine
Machine 150 is generated from the rotation of rotating shaft 160.Gas turbine 100 can be using natural gas, various types of
Synthesis gas and/or other types of fuel.
Controller 840 may include to be electrically coupled to each part and can be used in controlling the operation of each part
Any mechanism computerization control system.In a kind of infinite example, controller 840 can
With the controller that is GTG 100, (English name is:Turbine controller) etc..Pass
Sensor 820 or other monitoring devices may be coupled directly to the selected part in power plant 800, or can pass through
Controller 840 is electrically coupled to such part by other suitable interface mechanisms.Need herein
Bright is, although the quantity of sensor 820 is shown in Fig. 1 for three, but it is understood that, pass
The quantity of sensor 820 can be more than three.
Referring to Fig. 1, detecting fault processing unit 904 couples to realize event with control unit 840
Two-way communication between barrier detecting processing unit 904 and control unit 840.
Correct operation processing unit 906 couples to realize at correct operation with detecting fault processing unit 904
Two-way communication between reason unit 906 and detecting fault processing unit 904.In a kind of infinite implementation
In example, correct operation processing unit 206 is data acquisition and supervisor control (supervisory control
And data acquisition, SCADA), correct operation processing unit 206 can be located at the control in power plant 800
In room (not shown) processed.
Rule produces processing unit 902 to be coupled with power plant 800 by network 908.Network 908 can be right
Should be in any network type, including but not limited to dial-in network, practical network, PSTN
(PSTN), LAN (LAN), wide area network (WAN), Metropolitan Area Network (MAN) (MAN), personal area network
Network (PAN), virtual private net (VPN), campus network (CAN), storage area network (SAN),
Internet, in-house network or EtherType network, the network of two or more these type or other
Combination, with any kind of network topology in one or more the wiredly and/or wirelessly combination of communication link
Realize.In a kind of infinite embodiment, it is to calculate unit by multiple that rule produces processing unit 902
Into computer group or supercomputer.
Rule produces processing unit 902 for determining multiple association rule according to the historical data in power plant 800
Then.Specifically, rule produces processing unit 902 for the multiple previous failures in the historical data
Event and multiple previous alarm sequence determine multiple correlation rules.In a kind of infinite embodiment,
Each previous alarm sequence is the police produced in predetermined amount of time T before corresponding previous failures event occurs
Report sequence.
Detecting fault processing unit 904, for performing following operation:
Receive the current alert sequence in power plant 800;
Current alert sequence is compared with one or more the critical alert patterns in each correlation rule
Compared with;And
Produced when warning mode in current alert sequence is with one or more critical alert pattern match
Correspondence prediction failure for preventing event of failure generation in power plant;
Wherein, each critical alert pattern includes two related to the previous failures event in historical data
Or more than two alarms.
Above-mentioned current alert sequence includes the multiple alarms for arranging sequentially in time.Wherein, multiple sensings
Device 820 includes but is not limited to compressor inlet temperature sensor, compressor exhaust temperature sensor, compression
Machine inlet pressure transducer, compressor discharge pressure sensor, turbine inlet temperature sensor and turbine
Machine speed probe etc..For example, when the detecting temperature of compressor inlet temperature sensor is more than a pre- constant temperature
When spending, controller 840 produces alarm A;When the detecting temperature of compressor exhaust temperature sensor is more than one
During predetermined temperature, controller 840 produces alarm B.When the detecting temperature of turbine inlet temperature sensor
More than a predetermined temperature duration more than a predetermined amount of time when, controller 840 produces alarm C.
When the detecting pressure of compressor inlet pressure sensor is more than a predetermined pressure, controller 840 produces police
Report D.When the detecting pressure of compressor discharge pressure sensor is more than a predetermined pressure, controller 840
Produce alarm E etc..
In a kind of exemplary embodiment, the previous failures event in the historical data includes abnormal shutdown
(English name is trip event), normal shutdown (English name is shutdown event), critical alert
Or combinations thereof.In addition, above-mentioned abnormal shutdown also is understood as emergency cutoff.It is non-limiting in one kind
Example in, critical alert be turbine rotational speed sensor non-metering or all of turbine inlet temperature
The reading of sensor non-metering or all turbine inlet temperature sensors exceedes threshold values etc..
Correcting operation processing unit 906 is used to receive above-mentioned prediction failure and be performed according to the prediction failure to entangle
It is positive to operate to prevent the generation of event of failure in power plant.In a kind of exemplary embodiment, for preventing
The correction operation that only event of failure occurs in power plant includes the power-off operation in power plant, changes power plant
Reparation operation, the regulation dimension of software program in operation (for example, control power plant degraded running), power plant
The plan of repairing or combinations thereof etc..In other one exemplary embodiments, above-mentioned prediction failure can also be by
Notify to operator so that operator performed according to the prediction failure correct operation to prevent in power plant therefore
The generation of barrier event, such as operator may control power plant forced shutdown or maintenance or change power plant
Damage part in 800 etc..
Fig. 3 is referred to, a kind of method 300 of the offline determination correlation rule of its exemplary embodiment
Flow chart.Rule produces processing unit 902 to determine multiple correlation rules with offline for performing method 300,
The method 300 is comprised the following steps:
Step 302:Rule produces the multiple in the historical data in the acquisition of processing unit 902 power plant 800
Previous alarm sequence and multiple previous failures events.
Step 304:Rule produces processing unit 902 to know from multiple previous alarm sequences of historical data
Do not go out one or more event of failure samples.In an exemplary embodiment, one or more event of failures
Sample is any/any event of failure interested based on user's selection.
Step 306:For each event of failure sample, rule produces processing unit 902 to calculate and the event
The occurrence rate and misprediction rate of the related all warning modes of barrier event sample, and occurred according to multiple
Rate and multiple misprediction rates calculate corresponding multiple grades.In a kind of infinite embodiment, step
Warning mode in 306 is produced in predetermined amount of time T before the appearance of corresponding event of failure sample.
Step 308:Rule generation processing unit 902 will have calculating grade to be more than any one of threshold values
Warning mode is defined as critical alert pattern.
Step 310:Rule produces processing unit 902 to judge whether all of event of failure sample has been processed
FinishIf so, then flow returns to execution step 312;If it is not, then flow performs step 306.
Step 312:Rule produces processing unit 902 to determine multiple correlation rules, each correlation rule
Including one or more critical alert patterns and its grade and related to one or more critical alert patterns
Event of failure.
According to described by the above method 300, when all of event of failure sample process is finished, rule
Produce processing unit 902 to determine all of correlation rule, and all of correlation rule is sent to failure
Detecting processing unit 904 or power plant 800.
Fig. 4 is referred to, it is the method 400 of the online updating correlation rule of the first exemplary embodiment
Flow chart.Rule produces processing unit 902 for performing method 400 with the multiple association rule of online updating
Then, the method 400 is comprised the following steps:
Step 402:Rule produces processing unit 902 to receive the new event of failure and correspondence in power plant 800
Alarm.
Step 404:Rule produces whether processing unit 902 belongs to one or more event of failure samples
If so, then flow performs step 406;If it is not, then flow returns to execution step 402.
Step 406:Rule produces processing unit 902 to recognize the current alert related to the new event of failure
Pattern.In a kind of infinite embodiment, the current alert mode is gone out in corresponding new event of failure
Produced in predetermined amount of time T before now.
Step 408:When new event of failure belongs to event of failure sample, rule produces processing unit 902
The identification Historical Alerts pattern related to the new event of failure.In a kind of infinite embodiment, this is gone through
History warning mode is produced in predetermined amount of time T before the appearance of corresponding event of failure sample.
Step 410:Rule produces processing unit 902 to update all alarm moulds related to new event of failure
The frequency of occurrences of formula.
Step 412:Rule produces processing unit 902 according to multiple frequencies of occurrences for updating and multiple mistakes
Prediction rate updates corresponding multiple grades.
Step 414:Rule produces processing unit 902 any more than predetermined threshold with more New r4
One warning mode is defined as critical alert pattern.
Step 416:Rule produces processing unit 902 to update multiple correlation rules, each correlation rule bag
Include one or more critical alert patterns and its grade and related to one or more critical alert patterns
Event of failure.So far, flow is returned and performs step 402.
Fig. 5 is referred to, it is second method 500 of the online updating correlation rule of exemplary embodiment
Flow chart.Rule produces processing unit 902 for performing method 500 with the multiple association rule of online updating
Then, the method 500 is comprised the following steps:
Step 502:Rule produces processing unit 902 to receive prediction failure and the triggering in power plant 800
Predict one or more warning modes of failure.
Step 504:Rule produces processing unit 902 to judge whether the prediction failure is error predictionIf
It is that then flow performs step 506;If it is not, then flow returns to execution step 516.
Step 506:Rule produces processing unit 902 to increase the wrong pre- of one or more warning modes
Survey rate.
Step 516:Rule produces processing unit 902 to update the occurrence rate of one or more warning modes.
Step 508:Rule produces processing unit 902 according to multiple frequencies of occurrences for updating and multiple mistakes
Prediction rate updates corresponding multiple grades or according to multiple frequencies of occurrences and the misprediction rate of multiple increases
Update corresponding multiple grades;
Step 510:Rule produces processing unit 902 any more than predetermined threshold with more New r4
One warning mode is defined as critical alert pattern.
Step 512:Rule produces processing unit 902 to update multiple correlation rules, each correlation rule bag
Include one or more critical alert patterns and its grade and related to one or more critical alert patterns
Event of failure.So far, flow is returned and performs step 502.
Fig. 6 is referred to, it is a kind of flow chart of the failure prediction method 600 of exemplary embodiment.
Detecting fault processing unit 904 is used to perform failure prediction method 600 to produce the prediction in power plant 800
Failure, failure prediction method 600 is comprised the following steps:
Step 602:Detecting fault processing unit 904 obtains all of correlation rule, each correlation rule
Including one or more critical alert patterns and its grade and related to one or more critical alert patterns
Event of failure.
Step 604:Detecting fault processing unit 904 receives the current alert sequence in power plant 800.
Step 606:For one or more critical alert patterns, detecting fault in each correlation rule
Processing unit 904 judge warning mode in current alert sequence whether with one or more critical alerts
Pattern matchIf so, then flow performs step 608;If it is not, then flow returns to execution step 604.
Step 608:Detecting fault processing unit 904 produce power plant 800 correspondence prediction failure and its
Possibility.In a kind of one exemplary embodiment, the possibility of above-mentioned prediction failure is according to triggering prediction event
The grade of one or more warning modes of barrier is calculated.
Step 610:Detecting fault processing unit 904 judges whether all of correlation rule is disposed
If so, then flow performs step 604;If it is not, then flow returns to execution step 606.
The above-mentioned failure prediction system 900 and failure prediction method 600 of embodiment of the present invention, can be right
The historical data in power plant 800 is analyzed to determine multiple correlation rules, by the current of power plant 800
Alert sequence is compared with one or more the critical alert patterns in each correlation rule, and current
Power plant 800 is produced when warning mode in alert sequence is with one or more critical alert pattern match
Correspondence prediction failure.On the one hand, industry can be met to analysis power plant historical data and/or provides current
The demand of the significant comparing of data and historical data.On the other hand, above-mentioned prediction failure is notified to
Operator or correct operation processing unit 906 perform power-off operation, the fortune in modification power plant in power plant
Reparation operation, regulation maintenance project of software program etc. correct operation to prevent power plant in row, power plant
The generation of true fault event in 800;Therefore operator will not ignore above-mentioned prediction failure and correct behaviour
Make, the parts damages in power plant are so avoided to a certain extent and reduce power plant be forced close
The number of times of machine/non-predetermined shutdown.Namely improve the Mean Time Between Replacement (English of the pressure power-off in power plant 800
Literary fame is referred to as:Mean time between forced outage, MTBFO), and potentially add hair
The service life of part in power plant 800.
In addition, above-mentioned failure prediction system 900 and failure prediction method 600 also produce power plant 800
Prediction failure and its possibility, true fault is provided during this prevents power plant 800 to operator
Further aid/facility.
Although with reference to specific implementation method, the present invention is described, and those skilled in the art can
To understand, can be so that many modifications may be made and modification to the present invention.It is therefore contemplated that, claims
The all such modifications for being intended to be covered in true spirit of the present invention and scope and modification.
Claims (16)
1. a kind of failure prediction system, it includes:
Rule produces processing unit, for determining multiple correlation rules according to the historical data in power plant;And
Detecting fault processing unit, for performing following operation:
Receive the current alert sequence in power plant;
Current alert sequence is compared with one or more the critical alert patterns in each correlation rule
Compared with;And
Produced when warning mode in current alert sequence is with one or more critical alert pattern match
Correspondence prediction failure for preventing event of failure generation in power plant;
Wherein, each critical alert pattern includes two related to the previous failures event in historical data
Or more than two alarms.
2. failure prediction system as claimed in claim 1, it is characterised in that the failure prediction system is also
Including:
Operation processing unit is corrected, operation is corrected for receiving prediction failure and being performed according to the prediction failure
To prevent the generation of event of failure in power plant.
3. failure prediction system as claimed in claim 2, it is characterised in that:In for preventing power plant
The correction operation that event of failure occurs includes the power-off operation in power plant, the operation in modification power plant, generates electricity
The reparation operation of software program, regulation maintenance project or combinations thereof in factory.
4. failure prediction system as claimed in claim 1, it is characterised in that:Elder generation in the historical data
Prior fault event includes abnormal shutdown, normal shutdown, critical alert or combinations thereof.
5. failure prediction system as claimed in claim 1, it is characterised in that:The rule produces treatment single
The multiple previous failures events and multiple previous alarm sequences that unit is additionally operable in historical data are more to determine
Individual correlation rule.
6. failure prediction system as claimed in claim 5, it is characterised in that the rule produces treatment single
Unit is additionally operable to perform following operation to determine the plurality of correlation rule:
One or more event of failure samples are identified from all previous failures events of historical data;
Calculate the frequency of occurrences and error prediction of all warning modes related to each event of failure sample
Rate;
Corresponding multiple grades are calculated according to multiple frequencies of occurrences and multiple misprediction rates;And
There to be calculating grade to be defined as critical alert mould more than any one warning mode of predetermined threshold
Formula;Wherein, each warning mode includes two or more continuous alarms.
7. failure prediction system as claimed in claim 6, it is characterised in that:The rule produces treatment single
Unit is additionally operable to be determined when all event of failure sample process are finished multiple correlation rules and will determine
Multiple correlation rules are sent to power station or detecting fault processing unit, each correlation rule include one or
Multiple critical alert patterns and its grade and the failure thing related to one or more critical alert patterns
Part.
8. failure prediction system as claimed in claim 6, it is characterised in that:The rule produces treatment single
Unit is additionally operable to perform following operation to update the plurality of correlation rule:
Receive the new event of failure in power plant and corresponding alarm;
Recognized when new event of failure belongs to one or more event of failure samples related to the new event of failure
Current alert mode and Historical Alerts pattern;
Update the frequency of occurrences of all warning modes related to new event of failure;
Corresponding multiple grades are updated according to multiple frequencies of occurrences for updating and multiple misprediction rates;And
To be there is more New r4 to be defined as critical alert mould more than any one warning mode of predetermined threshold
Formula.
9. failure prediction system as claimed in claim 6, it is characterised in that:The rule produces treatment single
Unit is additionally operable to perform following operation to update the plurality of correlation rule:
Receive one or more warning modes of prediction failure and triggering prediction failure;
Increase the misprediction rate of one or more warning modes when predicting that failure is error prediction;
The occurrence rate of one or more warning modes is updated when predicting that failure is not error prediction;
Corresponding multiple grades or root are updated according to multiple frequencies of occurrences for updating and multiple misprediction rates
Corresponding multiple grades are updated according to the misprediction rate of multiple frequencies of occurrences and multiple increases;And
To be there is more New r4 to be defined as critical alert mould more than any one warning mode of predetermined threshold
Formula.
10. failure prediction system as claimed in claim 1, it is characterised in that:Error prediction treatment
Unit is additionally operable to warning mode and one or more critical alert pattern match in current alert sequence
When produce correspondence prediction failure and its possibility.
11. failure prediction systems as claimed in claim 10, it is characterised in that:Error prediction treatment
Unit is additionally operable to the grade according to one or more warning modes of triggering prediction failure to calculate prediction failure
Possibility.
A kind of 12. failure prediction methods, it includes:
Historical data according to power plant determines multiple correlation rules;
Receive the current alert sequence in power plant;
Current alert sequence is compared with one or more the critical alert patterns in each correlation rule
Compared with;And
Produced when warning mode in current alert sequence is with one or more critical alert pattern match
Correspondence prediction failure for preventing event of failure generation in power plant;
Wherein, each critical alert pattern includes two related to the previous failures event in historical data
Or more than two alarms.
13. failure prediction methods as claimed in claim 12, it is characterised in that the failure prediction method
Also include:
Receive prediction failure and performed according to the prediction failure and correct operation to prevent event of failure in power plant
Generation;Correction operation include power plant power-off operation, change the operation in power plant, in power plant
The reparation operation of software program, regulation maintenance project or combinations thereof.
14. failure prediction methods as claimed in claim 12, it is characterised in that determine the plurality of association
The method of rule includes:
One or more event of failure samples are identified from all previous failures events of historical data;
Calculate the frequency of occurrences and error prediction of all warning modes related to each event of failure sample
Rate;
Corresponding multiple grades are calculated according to multiple frequencies of occurrences and multiple misprediction rates;
There to be calculating grade to be defined as critical alert mould more than any one warning mode of predetermined threshold
Formula;And
Multiple correlation rules are determined when all event of failure sample process are finished;
Wherein, each warning mode includes two or more continuous alarms.
15. failure prediction methods as claimed in claim 14, it is characterised in that update the plurality of association
The method of rule includes:
Receive the new event of failure in power plant and corresponding alarm;
The identification current alert mode related to the new event of failure and Historical Alerts pattern;
The all alarm moulds related to new event of failure are updated when new event of failure belongs to event of failure sample
The frequency of occurrences of formula;
Corresponding multiple grades are updated according to multiple frequencies of occurrences for updating and multiple misprediction rates;And
To be there is more New r4 to be defined as critical alert mould more than any one warning mode of predetermined threshold
Formula.
16. failure prediction methods as claimed in claim 14, it is characterised in that update the plurality of association
The method of rule includes:
Receive one or more the alarm moulds in the current alert sequence of prediction failure and triggering prediction failure
Formula;
Increase the misprediction rate of one or more warning modes when predicting that failure is error prediction;
The occurrence rate of one or more warning modes is updated when predicting that failure is not error prediction;
Corresponding multiple grades or root are updated according to multiple frequencies of occurrences for updating and multiple misprediction rates
Corresponding multiple grades are updated according to the misprediction rate of multiple frequencies of occurrences and multiple increases;And
To be there is more New r4 to be defined as critical alert mould more than any one warning mode of predetermined threshold
Formula.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510837616.9A CN106802643A (en) | 2015-11-26 | 2015-11-26 | failure prediction system and method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510837616.9A CN106802643A (en) | 2015-11-26 | 2015-11-26 | failure prediction system and method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN106802643A true CN106802643A (en) | 2017-06-06 |
Family
ID=58976635
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510837616.9A Pending CN106802643A (en) | 2015-11-26 | 2015-11-26 | failure prediction system and method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106802643A (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108228800A (en) * | 2017-12-29 | 2018-06-29 | 无锡十新能源投资有限公司 | A kind of photovoltaic generating system abnormality detection system and method based on data mining |
CN110672332A (en) * | 2019-09-10 | 2020-01-10 | 上海电力大学 | Gas turbine fault early warning system based on SARIMA model |
CN111712771A (en) * | 2017-11-10 | 2020-09-25 | Abb瑞士股份有限公司 | Data processing apparatus and method capable of performing problem diagnosis in production system having a plurality of robots |
CN112488181A (en) * | 2020-11-26 | 2021-03-12 | 哈尔滨工程大学 | Service fault high-response matching method based on MIDS-Tree |
CN113544486A (en) * | 2019-03-08 | 2021-10-22 | 三菱电机株式会社 | Fault diagnosis system, prediction rule generation method, and prediction rule generation program |
CN114641740A (en) * | 2019-11-05 | 2022-06-17 | Abb瑞士股份有限公司 | Method and device for monitoring an electric drive in an industrial system |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1656599A (en) * | 2002-05-29 | 2005-08-17 | 东京电子株式会社 | Method and apparatus for monitoring tool performance |
CN1655325A (en) * | 2003-12-26 | 2005-08-17 | 佳能系统集成有限公司 | Device for and method of creating a model for determining relationship between process and quality |
CN1761919A (en) * | 2003-03-18 | 2006-04-19 | 费舍-柔斯芒特系统股份有限公司 | Asset optimization reporting in a process plant |
CN101799359A (en) * | 2010-01-27 | 2010-08-11 | 北京信息科技大学 | Failure monitoring and predicting method and system of power equipment |
CN102193836A (en) * | 2011-04-18 | 2011-09-21 | 电子科技大学 | Dynamic preventative maintenance method for electromechanical equipment |
CN102830666A (en) * | 2011-05-11 | 2012-12-19 | 通用电气公司 | System and method for optimizing plant operations |
-
2015
- 2015-11-26 CN CN201510837616.9A patent/CN106802643A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1656599A (en) * | 2002-05-29 | 2005-08-17 | 东京电子株式会社 | Method and apparatus for monitoring tool performance |
CN1761919A (en) * | 2003-03-18 | 2006-04-19 | 费舍-柔斯芒特系统股份有限公司 | Asset optimization reporting in a process plant |
CN1655325A (en) * | 2003-12-26 | 2005-08-17 | 佳能系统集成有限公司 | Device for and method of creating a model for determining relationship between process and quality |
CN101799359A (en) * | 2010-01-27 | 2010-08-11 | 北京信息科技大学 | Failure monitoring and predicting method and system of power equipment |
CN102193836A (en) * | 2011-04-18 | 2011-09-21 | 电子科技大学 | Dynamic preventative maintenance method for electromechanical equipment |
CN102830666A (en) * | 2011-05-11 | 2012-12-19 | 通用电气公司 | System and method for optimizing plant operations |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111712771A (en) * | 2017-11-10 | 2020-09-25 | Abb瑞士股份有限公司 | Data processing apparatus and method capable of performing problem diagnosis in production system having a plurality of robots |
US11669081B2 (en) | 2017-11-10 | 2023-06-06 | Abb Schweiz Ag | Data processing device capable of performing problem diagnosis in a production system with plurality of robots and method |
CN111712771B (en) * | 2017-11-10 | 2024-03-12 | Abb瑞士股份有限公司 | Data processing apparatus and method capable of performing problem diagnosis |
CN108228800A (en) * | 2017-12-29 | 2018-06-29 | 无锡十新能源投资有限公司 | A kind of photovoltaic generating system abnormality detection system and method based on data mining |
CN108228800B (en) * | 2017-12-29 | 2022-03-15 | 无锡十一新能源投资有限公司 | Photovoltaic power generation system anomaly detection system and method based on data mining |
CN113544486A (en) * | 2019-03-08 | 2021-10-22 | 三菱电机株式会社 | Fault diagnosis system, prediction rule generation method, and prediction rule generation program |
CN113544486B (en) * | 2019-03-08 | 2022-10-28 | 三菱电机株式会社 | Fault diagnosis system, fault prediction method, and computer-readable recording medium |
CN110672332A (en) * | 2019-09-10 | 2020-01-10 | 上海电力大学 | Gas turbine fault early warning system based on SARIMA model |
CN114641740A (en) * | 2019-11-05 | 2022-06-17 | Abb瑞士股份有限公司 | Method and device for monitoring an electric drive in an industrial system |
CN112488181A (en) * | 2020-11-26 | 2021-03-12 | 哈尔滨工程大学 | Service fault high-response matching method based on MIDS-Tree |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106802643A (en) | failure prediction system and method | |
US9360864B2 (en) | Turbine fault prediction | |
JP6250942B2 (en) | System and method for operation with improved reliability | |
CN102830666B (en) | Optimize the system and method for plant operation | |
EP2820490B1 (en) | System for condition monitoring of a group of plants | |
EP2241726A2 (en) | A method for the repair of an engine and corresponding systems for monitoring this engine | |
US9280617B2 (en) | Systems and methods for improved reliability operations | |
EP2685335A2 (en) | System and method for monitoring process control system health | |
CN110431502A (en) | Omen detection system and omen detection method | |
JP2005317002A (en) | Method and system for monitoring machinery | |
EP3055746A1 (en) | Correlation and annotation of time series data sequences to extracted or existing discrete data | |
CN113705924A (en) | Intelligent diagnosis method and system for thermal control equipment | |
EP3945383A1 (en) | Management of a filter in continuous flow engine | |
CN116483054A (en) | Industrial robot running state monitoring and early warning system and method | |
EP3303835B1 (en) | Method for windmill farm monitoring | |
CN116235121A (en) | Apparatus and method for identifying anomalies in industrial facilities for performing a production process | |
CN109520740A (en) | Method for forecasting gas turbine health | |
Stefanowicz et al. | The method of determination and detection of abnormal states in a GT8C turbine | |
CN114692990A (en) | Hydroelectric generating set fault early warning and prejudging method based on Aspen Mtell system | |
CN117573995A (en) | Method for displaying equipment state data in production process | |
CN117289659A (en) | Intelligent automatic monitoring system for centralized control operation of power plant | |
Miyoshi et al. | Hybrid Monitoring of Offshore Compression Systems | |
Mochi et al. | Improved Oil & Gas Plant Operations Through Advanced Remote Services Driven By Operating Data |
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 | ||
AD01 | Patent right deemed abandoned | ||
AD01 | Patent right deemed abandoned |
Effective date of abandoning: 20200811 |