CN103471729A - Device temperature early warning method and application thereof - Google Patents

Abstract

The invention provides a device temperature early warning method and an application thereof. The method comprises the following steps: conducting data collection on the temperature of a measured device, and storing the temperature data into a database; comparing the collected data with the set limiting values, if the collected data are larger than the limiting values, giving an alarm, and if the collected data are smaller than the limiting values, conducting the next step; conducting data analysis after one cycle of the data collection is completed; respectively obtaining the difference values between the obtained sets of temperature data, comparing the difference values of the sets of data with the limiting values, if the difference values exceed the difference value limiting values, giving an alarm, and if the difference values do not exceed the difference value limiting values, conducting the next step; calculating the deviation between the theoretical temperature value and the actual temperature value, the mean square error of the deviation and correlation coefficients of the deviation according to the collected data in the one cycle; judging the status of the measured device according to the mean square error, the correlation coefficients and the deviation. According to the device temperature early warning method and the application of the method, components can be prevented from working in the abnormal status for a long time, corresponding maintenance is timely carried out on a unit, consumption of spare parts is reduced, and the working efficiency is improved.

Description

A kind of unit temp method for early warning and application thereof

Technical field

The present invention relates to device temperature early warning field, particularly relate to a kind of unit temp method for early warning and application thereof.

Background technology

Involving great expense of main equipment, the cost of systematic part is also higher, and if there is damage, repair time is long, affects its normal operation.Take the high-power wind-driven generator group as example, if this situation occurs in the reasonable time of wind speed, will greatly affect generated energy, directly cause the economic loss of electricity power enterprise.If can be before fault occurs, the potential faults existed in discovering device, carry out suitable maintenance, and parts will be effectively protected, and can greatly reduce the damage of parts, improves parts serviceable life; To the damage of those unrepairables, also be convenient to be ready in advance parts, reduce stop time.

In traditional wind power generating set is safeguarded, after many employing faults occur, failure cause is analyzed, and then the flow process of repairing, the early warning of shortage to the unit duty, many times, when the situations such as temperature anomaly appear in the parts of unit, can not find in time, parts are gone to work braving one's illness, finally cause the damage of original paper, increased maintenance cost and maintenance time, reduced the generating efficiency of blower fan.Along with the installed capacity of Large-scale Wind Turbines is constantly soaring, the cost control of blower fan manufacturer and electricity power enterprise are more and more higher to the requirement of generating efficiency, to the temperature pre-warning of wind power generating set main parts size, are very necessary.

At present the temperature of wind power generating set main parts size do not had to concrete method for early warning, the form that just adopts threshold values to report to the police, set a threshold values, when temperature surpasses threshold values, just starts to report to the police, and its workflow as shown in Figure 1.

The shortcoming that existing method exists mainly contain following some:

1, only have one to set threshold values, when the main parts long-term work of wind power generating set is being no more than threshold values but at improper temperature the time, can not be being found in time.

2, only individual data point is judged, easily cause unnecessary wrong report or fail to report, the accuracy of early warning is low.

As can be seen here, above-mentioned existing a kind of unit temp method for early warning in the use, still has inconvenience and defect, and urgently is further improved.How to found a kind of new unit temp method for early warning that can find in time parts temperature anomaly, improve plant factor, becoming the current industry utmost point needs improved target.

Summary of the invention

The technical problem to be solved in the present invention is to provide a kind of unit temp method for early warning and application thereof, can find in time that parts are abnormal, safeguard in time to improve part life, thereby overcome the deficiency that existing temperature pre-warning accuracy is low, affect the equipment work efficiency.

For solving the problems of the technologies described above, the invention provides a kind of unit temp method for early warning, comprise the following steps: A. is gathered the temperature of surveyed device; B. the data after gathering are stored in database; C. the limit value of the data of collection and setting is compared, give the alarm if be greater than limit value, if be less than limit value, carry out step D; D. judge that whether data collection cycle is complete, data acquisition enters step e after meeting one-period, does not meet and returns to steps A; E. do poorly to the temperature data of respectively organizing gathered respectively, the difference of each group data is compared with the difference limit value of setting, if surpass the difference limit value, send warning, if over the difference limit value enter step F; F. calculate the deviation between theoretical temperature value and actual temperature value, mean square deviation and the related coefficient of deviation according to the data in the one-period collected; G. according to mean square deviation, related coefficient and deviation, the state of surveyed device is judged.

As a kind of improvement, step F is to calculate theoretical temperature value by the off-line data temperature model.

The mean square deviation of described deviation, deviation and the computing method of related coefficient are as follows:

DEV＝P_DATA-R_DATA

$\mathrm{MSE}=\sqrt{\frac{1}{N}\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{(P\_\mathrm{DATA}\left(i\right)-R\_\mathrm{DATA}\left(i\right))}^2}$

$\mathrm{SCC}=\frac{N\underset{i=1}{\overset{N}{\mathrm{\Σ}}}P\_\mathrm{DATA}\left(i\right)\×R\_\mathrm{DATA}\left(i\right)-\underset{i=1}{\overset{N}{\mathrm{\Σ}}}P\_\mathrm{DATA}\left(i\right)\×\underset{i=1}{\overset{N}{\mathrm{\Σ}}}R\_\mathrm{DATA}\left(i\right)}{\sqrt{N\underset{i=1}{\overset{N}{\mathrm{\Σ}}}P\_\mathrm{DATA}{\left(i\right)}^2-{(\underset{i=1}{\overset{N}{\mathrm{\Σ}}}P\_\mathrm{DATA}\left(i\right))}^2}\×\sqrt{N\underset{i=1}{\overset{N}{\mathrm{\Σ}}}R\_\mathrm{DATA}{\left(i\right)}^2-{(\underset{i=1}{\overset{N}{\mathrm{\Σ}}}R\_\mathrm{DATA}\left(i\right))}^2}}$

Wherein, P_DATA is the theoretical temperatures value, and R_DATA is actual temperature value, and DEV is deviation, the mean square deviation that MSE is deviation, and SCC is related coefficient, N is the data length in a collection period.

If the deviation when becoming the oar motor and normally moving is deviation standard value S, described step G is specially: if mean square deviation, related coefficient all lower than warning line, and without any the deviation of data higher than 1S, output state grade 1; If mean square deviation, related coefficient be all lower than warning line, and the deviation that a data point is arranged is between 1S and 2S, output state grade 2; If mean square deviation, related coefficient be all lower than warning line, and the deviation that two data points are arranged is between 1S and 2S, and these two points are discontinuous, output state grade 3.If mean square deviation, related coefficient be all lower than warning line, and the deviation that a data point is arranged is between 2S and 3S, the output state class 4; If mean square deviation, related coefficient be all lower than warning line, and the deviation that two consecutive numbers strong points are arranged is between 1S and 2S, the output state class 5; If mean square deviation, related coefficient be all lower than warning line, and the deviation that three data points are arranged is between 1S and 2S, the output state class 6; If mean square deviation, related coefficient be all lower than warning line, and the deviation that the data point more than four is arranged is between 1S and 2S, output state grade 7; If mean square deviation, related coefficient be all lower than warning line, and the deviation that two or three data points are arranged is between 2S and 3S, output state grade 8; If any one surpasses warning line mean square deviation, related coefficient, or have the deviation of any one data to surpass 3S, or the deviation that 4 above data are arranged is between 2S and 3S, output state grade 9.

When described state grade is 4,5,6, notice that parts data changes, if data mode worsens, is checked parts; When described state grade is 7,8,9, gives more sustained attention parts data and change, and parts are checked.

The present invention also provides the change oar motor of a kind of said method in wind power generating set, becomes the oar inverter, becomes the application on oar standby power supply, IGBT rectification unit, IGBT inversion unit, converter system reactor and/or generator windings.

Further, sensor installation outside described pitch-controlled system, gathered temperature, voltage, electric current, current wind speed, the environment temperature that becomes the oar motor.

After adopting such design, the present invention at least has the following advantages:

1, the present invention is applicable to temperature monitoring and the early warning of main equipment, is particularly useful in wind power generating set, and maintenance test that can be in good time to blower fan, improved the generating efficiency of aerogenerator, reduced maintenance cost.

2, the present invention has carried out the analysis of different levels to the data of wind power generating set main parts size, many early warning lines have been set up, the form that adopts different piece mutually to contrast, utilize advanced intelligent algorithm, the problem from different perspectives unit existed is judged, take full advantage of the service data of main parts size, can find in time that the wind power generating set main parts size is in operation abnormal, can prevent critical piece to work long hours in abnormal state of affairs, before problem occurs, remind maintainer's in good time (as when wind speed is less) to safeguard accordingly unit, reduced the consumption of spare part, reduce the loss of generated energy.

The accompanying drawing explanation

Above-mentioned is only the general introduction of technical solution of the present invention, and in order to better understand technological means of the present invention, below in conjunction with accompanying drawing and embodiment, the present invention is described in further detail.

Fig. 1 is existing methodical temperature pre-warning process flow diagram.

Fig. 2 is unit temp method for early warning process flow diagram of the present invention.

Fig. 3 is the process flow diagram of unit temp method for early warning data prediction part of the present invention.

Embodiment

The change oar motor temperature early warning of wind generating set pitch control system of below take is example, and this programme is explained in detail.Unit temp method for early warning provided by the present invention comprises the following steps:

A. sensor installation outside pitch-controlled system, gathered temperature and other operational datas that becomes the oar motor; B. the data after gathering are stored in database; C. the limit value of the data of collection and setting is compared, give the alarm if be greater than limit value, if be less than limit value, carry out step D; D. judge that whether data collection cycle is complete, data acquisition enters step e after meeting one-period, does not meet and returns to steps A; E. it is poor that the temperature data that respectively each group is become to the oar motor is done, and the difference of each group data and the difference limit value of setting are compared, if surpass the difference limit value, sends warning, if do not surpass the difference limit value enter next step data prediction part; F. according to the data in the one-period that collects calculate theoretical temperature value and and actual temperature value between deviation, mean square deviation and the related coefficient of deviation; G. according to mean square deviation, related coefficient and deviation, the state that becomes the oar motor is judged.

Particularly, this unit temp method for early warning is further elaborated, establishing described pitch-controlled system has three groups, will become the data such as the temperature, voltage, electric current of oar motor and current wind speed, environment temperature by the sensor of pitch-controlled system periphery and be gathered and process.The data that obtain are stored in database, judge whether severely subnormal of data, with the limit value of prior setting relatively.If be greater than limit value, quote immediately alarm, if be less than limit value, judge whether data collection cycle completes.Limit value is that the user can accept to set in scope in system, such as temperature range-30-200, as the user thinks, becomes the oar motor temperature higher than 100 degree on the component life impact greatly, limit value can be set as to 100.If, after the full one-period of data acquisition, after reaching the data volume that meets data analysis, enter the data analysis part.The data centralized procurement cycle can be set as required, usually at 30 minutes, arbitrarily sets in by 10 hours, and for example 30 minutes, 1 hour or 2 hours, the enough analysis component running statuses of data that gather in each cycle got final product.

Data analysis is partly the core that becomes the early warning of oar motor, comprises difference comparison, data prediction part and state estimation part.At first it is poor the temperature data of three change oar motors collection to be done respectively, establishes three groups of change oar motor temperatures and is respectively T1, T2, T3, and the difference of three temperature is respectively SUB1, SUB2, SUB3, has

SUB1=T1-T2

SUB2＝T1-T3

SUB3＝T2-T3

Judge whether SUB1, SUB2, SUB3 surpass the difference limit value of setting, if surpass, are directly reported to the police, if do not surpass, enter next step data prediction part.The difference limit value is rule of thumb set usually, if temperature surpasses the difference limit value, now has the oar of change machine operation abnormal, very large on becoming the impact of oar electrical machinery life, needs the timely Inspection and maintenance of staff.

Data prediction partly adopts the method for continuous time series prediction, and this part is calculated and formed by off-line data modeling and on-line prediction.

1) off-line data modeling: by the qualitative analysis to off-line data, find out change oar motor operating voltage, motor working current, wind speed and environment temperature variable that impact becomes the oar motor temperature, utilize data validity, first data are carried out to pre-service, retain valid data, because these data are time series datas, therefore according to " 80/20 rule ", divide training set and test set data, adopt linear (Linear) kernel function K (x, x _i)=xx _isupport vector machine (SVM) algorithm, wherein: x _ithe input variable sequences such as change oar motor operating voltage, motor working current, wind speed and environment temperature that the impact chosen becomes the oar motor temperature, set up and become oar motor temperature model, and by penalty coefficient and the kernel functional parameter of genetic algorithm optimization SVM, finally by error evaluation, assess the size of square error and absolute error, thereby estimate the modeling effect of whole model.

2) on-line prediction calculates: the data such as temperature, voltage, electric current and the wind speed of the change oar motor of the one-period that part of data acquisition is gathered, environment temperature, be input in the model established in advance, according to known parameters, all data to this cycle are predicted output, and, using the predicted value of the change oar motor temperature exported in model as the theoretical temperatures value P_DATA that becomes the oar motor, calculate mean square deviation MSE and the related coefficient SCC of deviation D EV, the deviation D EV of theoretical temperature value P_DATA and actual temperature value R_DATA.If an interior data length of collection period is N, have

DEV＝P_DATA-R_DATA （1）

$\mathrm{MSE}=\sqrt{\frac{1}{N}\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{(P\_\mathrm{DATA}\left(i\right)-R\_\mathrm{DATA}\left(i\right))}^2}---\left(2\right)$

$\mathrm{SCC}=\frac{N\underset{i=1}{\overset{N}{\mathrm{\Σ}}}P\_\mathrm{DATA}\left(i\right)\×R\_\mathrm{DATA}\left(i\right)-\underset{i=1}{\overset{N}{\mathrm{\Σ}}}P\_\mathrm{DATA}\left(i\right)\×\underset{i=1}{\overset{N}{\mathrm{\Σ}}}R\_\mathrm{DATA}\left(i\right)}{\sqrt{N\underset{i=1}{\overset{N}{\mathrm{\Σ}}}P\_\mathrm{DATA}{\left(i\right)}^2-{(\underset{i=1}{\overset{N}{\mathrm{\Σ}}}P\_\mathrm{DATA}\left(i\right))}^2}\×\sqrt{N\underset{i=1}{\overset{N}{\mathrm{\Σ}}}R\_\mathrm{DATA}{\left(i\right)}^2-{(\underset{i=1}{\overset{N}{\mathrm{\Σ}}}R\_\mathrm{DATA}\left(i\right))}^2}}---\left(3\right)$

According to the size of mean square deviation MSE and related coefficient SCC, and the Changing Pattern of N deviation D EV in data collection cycle, by the data mining mode, set up the data evaluation rule, the health status that becomes the oar motor is carried out to data assessment.The health that will become the oar motor in the present embodiment is divided for nine grades, and the parameter that data prediction is partly obtained is input to the data assessment part, finally obtains the health status of blower fan.

In the present invention, the concrete evaluation method of data assessment part is as follows:

According to wind power generating set running state data in the past, the mean square deviation MSE when setting unit and need to safeguard and the value of related coefficient SCC are warning line, and the deviation of establishing when becoming the oar motor and normally moving is deviation standard value S.The size of warning line and deviation standard value S is according to design and operation characteristic and the protection feature requirement of distinct device, and the service data statistics of bonding apparatus is determined.The overcurrent protection limit value of supposing device A is H, and device A belongs to rapid wear high value parts, and warning line and deviation standard value are little.Otherwise, can, in the scope of meeting design requirement, suitably relax warning line and deviation standard value.

(1) mean square deviation MSE, related coefficient SCC be lower than warning line, without any the DEV of data higher than 1S, output state grade 1.

(2) mean square deviation MSE, related coefficient SCC, lower than warning line, have the DEV of a data point between 1S and 2S, output state grade 2.

(3) mean square deviation MSE, related coefficient SCC, lower than warning line, have the DEV of two data points between 1S and 2S, and these two points are discontinuous, output state grade 3.

(4) mean square deviation MSE, related coefficient SCC, lower than warning line, have the DEV of a data point between 2S and 3S, the output state class 4.

(5) mean square deviation MSE, related coefficient SCC, lower than warning line, have the DEV at two consecutive numbers strong points between 1S and 2S, the output state class 5.

(6) mean square deviation MSE, related coefficient SCC, lower than warning line, have the DEV of three points between 1S and 2S, the output state class 6.

(7) mean square deviation MSE, related coefficient SCC, lower than warning line, have the DEV of the point more than four between 1S and 2S, output state grade 7.

(8) mean square deviation MSE, related coefficient SCC, lower than warning line, have the DEV of two or three points between 2S and 3S, output state grade 8.

(9) if mean square deviation MSE, related coefficient SCC any one surpass warning line, or have the deviation D EV of any one data to surpass 3S, or the DEV that 4 above data are arranged is between 2S and 3S, output state grade 9;

State grade 1-3 shows that the parts state, for " excellent ", illustrates that the parts running status is fine, without carrying out any operation;

State grade 4-6 shows that the parts state, for " good ", illustrates that the parts running status is good, needs to pay close attention to the variation of parts data, if data mode worsens, needs to send someone parts are checked;

State grade 7-9 shows that the parts state, for " poor ", illustrates that the parts running status is bad, has hidden danger, need to give more sustained attention parts data and change, and send someone parts are checked, removes a hidden danger;

Every kind of state grade is divided into again 3 little ranks, and the degree of expression state quality and the concern of every kind of state grade check priority, and numeral is larger, and state is poorer, pays close attention to and checks that priority is higher.Such as the evaluation status that 3 parts are arranged is respectively 7,8,9,3 unit statuss are all " poor ", all need to send someone to be checked, but, in the situation that personnel are limited, the parts that the priority check state is 9.

The pitch-controlled system that unit temp method for early warning involved in the present invention mainly is applicable to Large-scale Wind Turbines (comprises and becomes the oar motor, become the oar inverter, become oar standby power supply etc.), converter system (comprising IGBT rectification unit, IGBT inversion unit, reactor etc.) and generator windings.

The above; it is only preferred embodiment of the present invention; not the present invention is done to any pro forma restriction, those skilled in the art utilize the technology contents of above-mentioned announcement to make a little simple modification, equivalent variations or modification, all drop in protection scope of the present invention.

Claims (7)

1. a unit temp method for early warning is characterized in that comprising the following steps:

A. the temperature data of surveyed device gathered;

B. the data after gathering are stored in database;

C. the limit value of the data of collection and setting is compared, give the alarm if be greater than limit value, if be less than limit value, carry out step D;

D. judge that whether data collection cycle is complete, data acquisition enters step e after meeting one-period, does not meet and returns to steps A;

E. do poorly to the temperature data of respectively organizing gathered respectively, the difference of each group data is compared with the difference limit value of setting, if surpass the difference limit value, send warning, if over the difference limit value enter step F;

F. according to the data in the one-period that collects calculate theoretical temperature value and and actual temperature value between deviation, mean square deviation and the related coefficient of deviation;

G. according to mean square deviation, related coefficient and deviation, the state of surveyed device is judged.

2. a kind of unit temp method for early warning according to claim 1, it is characterized in that: described step F is to calculate theoretical temperature value by the off-line data temperature model.

3. a kind of unit temp method for early warning according to claim 1 is characterized in that the computing method of the mean square deviation of described deviation, deviation and related coefficient are as follows:

DEV＝P_DATA-R_DATA

$\mathrm{MSE}=\sqrt{\frac{1}{N}\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{(P\_\mathrm{DATA}\left(i\right)-R\_\mathrm{DATA}\left(i\right))}^2}$

$\mathrm{SCC}=\frac{N\underset{i=1}{\overset{N}{\mathrm{\Σ}}}P\_\mathrm{DATA}\left(i\right)\×R\_\mathrm{DATA}\left(i\right)-\underset{i=1}{\overset{N}{\mathrm{\Σ}}}P\_\mathrm{DATA}\left(i\right)\×\underset{i=1}{\overset{N}{\mathrm{\Σ}}}R\_\mathrm{DATA}\left(i\right)}{\sqrt{N\underset{i=1}{\overset{N}{\mathrm{\Σ}}}P\_\mathrm{DATA}{\left(i\right)}^2-{(\underset{i=1}{\overset{N}{\mathrm{\Σ}}}P\_\mathrm{DATA}\left(i\right))}^2}\×\sqrt{N\underset{i=1}{\overset{N}{\mathrm{\Σ}}}R\_\mathrm{DATA}{\left(i\right)}^2-{(\underset{i=1}{\overset{N}{\mathrm{\Σ}}}R\_\mathrm{DATA}\left(i\right))}^2}}$

Wherein, P_DATA is the theoretical temperatures value, and R_DATA is actual temperature value, and DEV is deviation, the mean square deviation that MSE is deviation, and SCC is related coefficient, N is the data length in a collection period.

4. a kind of unit temp method for early warning according to claim 1, is characterized in that, the deviation of establishing when becoming the oar motor and normally moving is deviation standard value S, and described step G is specially:

If mean square deviation, related coefficient be all lower than warning line, and without any the deviation of data higher than 1S, output state grade 1;

If mean square deviation, related coefficient be all lower than warning line, and the deviation that a data point is arranged is between 1S and 2S, output state grade 2;

If mean square deviation, related coefficient be all lower than warning line, and the deviation that two data points are arranged is between 1S and 2S, and these two points are discontinuous, output state grade 3.

If mean square deviation, related coefficient be all lower than warning line, and the deviation that a data point is arranged is between 2S and 3S, the output state class 4;

If mean square deviation, related coefficient be all lower than warning line, and the deviation that two consecutive numbers strong points are arranged is between 1S and 2S, the output state class 5;

If mean square deviation, related coefficient be all lower than warning line, and the deviation that three data points are arranged is between 1S and 2S, the output state class 6;

If mean square deviation, related coefficient be all lower than warning line, and the deviation that the data point more than four is arranged is between 1S and 2S, output state grade 7;

If mean square deviation, related coefficient be all lower than warning line, and the deviation that two or three data points are arranged is between 2S and 3S, output state grade 8;

If any one surpasses warning line mean square deviation, related coefficient, or have the deviation of any one data to surpass 3S, or the deviation that 4 above data are arranged is between 2S and 3S, output state grade 9.

5. a kind of unit temp method for early warning according to claim 4 is characterized in that:

When described state grade is 4,5,6, notice that parts data changes, if data mode worsens, is checked parts;

When described state grade is 7,8,9, gives more sustained attention parts data and change, and parts are checked.

In a claim 1-5 the described method of any one at the change oar motor of wind power generating set, become the oar inverter, become the application on oar standby power supply, IGBT rectification unit, IGBT inversion unit, converter system reactor and/or generator windings.

7. the application of the described method of any one on the change oar motor of wind generating set pitch control system in a claim 1-5, it is characterized in that sensor installation outside described pitch-controlled system, temperature, voltage, electric current, current wind speed, the environment temperature that becomes the oar motor gathered.

Images