Dynamic prediction of CH4 emission in longwalls

Christian Tauziede, Zbigniew Pokryszka

To cite this version:

Christian Tauziede, Zbigniew Pokryszka. Dynamic prediction of CH4 emission in longwalls.25. Conference Internationale des Instituts de Recherches sur la Securite dans les Mines, Sep1993, Pretoria, South Africa. pp.41-50, 1993. <ineris-00971873>

HAL Id: ineris-00971873

https://hal-ineris.ccsd.cnrs.fr/ineris-00971873

Submitted on 3 Apr 2014

HAL is a multi-disciplinary open accessarchive for the deposit and dissemination of sci-entific research documents, whether they are pub-lished or not. The documents may come fromteaching and research institutions in France orabroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, estdestinee au depot et a la diffusion de documentsscientifiques de niveau recherche, publies ou non,emanant des etablissements d’enseignement et derecherche francais ou etrangers, des laboratoirespublics ou prives.

brought to you by COREView metadata, citation and similar papers at core.ac.uk

provided by HAL-INERIS

DYNAMIC PREDICTION OF CH4 EMISSION IN LONGWALLS

Christian TAUZffiDE and Zbigniew POKRYSZKA

INERISVemeuil en Halatte

France

SUMMARY

Methods for predicting mean CH4emission in longwalls have beendevelopped in France and other countriesfor many years. These methods have beenwidely validated but remain limited intheir use because of the continuousincrease in productivity of such working.

Therefore, development of dynamicprediction methods - e.g. allowingprediction of emission on a daily orweekly basis - is strongly necessary.

Aiming to that, two ways have beeninvestigated :

- a statistical approach. Analysis of CH4emission of 14 faces has allowedbuilding the expression of the emissionduring a given week äs a function ofmean predicted emission and of advance(or production) of the three previousweeks.

- a mathematical modelling of CH4emission äs a function of stratigraphyand of gas content of the stratasurrounding the mined seam. Analyticalexpression contains numerousParameters that must be adjusted using areference face, so äs to be able to carryout a real prediction for later workings.

INTRODUCTION

Prediction methods have been largelytried and tested for a number of years butremain limited due to the increase in theproductivity of workings. The advancerate of longwalls can be very high, whichhas two consequences:

• the time required for the emission tobuild up at the beginning of the panelis sometimes relatively high,

• the variations in emission with regardto the average are also important.

The methods used to predict the averageemission are not sufficient to account forthese two phenomena (see Figure l).Dynamic prediction methods thereforeneed to be developed which will enablethe day-by-day or week-by-weekemission to be ascertained äs a functionof the activity of the previous periods.Most of the authors who have examinedthe problem use a statistical approach topredict firedamp emission. They proposetaking the average, the dispersion and thefrequency of the emission values äs thecharacteristic parameters of thisphenomenon (Bruyet, 1967; Borowski,1969; Winter, 1972).

This approach enables the "normal"emission level to be defined, äs well äs asafety factor resulting from the

25eme Conference Internationale des Instituts de Recherches sur la Securite dans les Mines,Pretoria, 13-17 septembre 1993, vol. 5, p. 27-40.

probability of deviating from this level.However, it does not enable a realprediction of the emission to be made äs afunction of the activity.

Another approach which contains acertain amount of dynamic prediction hasalready been proposed (Kaffanke, 1980).Kaffanke expresses the firedampemission for each day of the week äs acertain combination of several factors(production of the day in question,cumulated production of the previousdays). The functions used to caiculate thedaily emission are different for each dayof the week. The factors proposed äshaving an influence on the emission offiredamp are not related to the physicalmechanisms. The success of the Kaffankemodel therefore rests, to a large extent, onthe low variability of the weeklyproduction cycles.

An approach has also been proposedbased on the hypothesis that the gasemission during a given day depends onthe production of that day and on that of acertain number of previous days (BritishCoal, 1988). In this model, the functionexpressing the emission is applied in thesame way to all the days of the week.

The constants of this function arecaiculated for a given working accordingto the emission values measured in thisvery working. Once the constants havebeen determined, the firedamp emissionis predicted for the remaining part of theworking.

METHOD USED IN FRANCE TOPREDIGT THE AVERAGE EMISSION

IN LONGWALLS

The method widely used in France forpredicting firedamp emission in faces wasdeveloped more than 25 ye'ars ago(Günther, 1965). It has been improved on

several occasions, and particularly inrecent times (Jeger, 1980; Ineris, 1992).This model is based on severalfundamental observations :

• The emission of firedamp is the resultof partial degassing of the coal seamsand rock beds situated in a volume ofinfluence at the roof and the floor öfthe seam being mined. After a certainadvance of the face, corresponding tothe extension of the volume ofinfluence, it becomes stabilized (seeFigure 2). The limits of the volumedegassed by a face are in the order of170 meters for the roof and 60 metersfor the floor of the mined seam.

• In the volume of influence, thedesorbable gas content of a seamdecreases from its initial content to afinal residual content. This residualcontent depends on the distance of theseam being considered from the seambeing mined. Figure 3 indicates thedegassing rates of the affected seams.For the mined seam, a degassing rateof 50 % is usually taken.

• In the volume of influence, the rockbeds also release part of the firedampinitially compressed in theirintergranular voids. The amount ofgas released is caiculated according tothe porosity of the rocks, the initial insitu pressure and the residual pressure(Ineris, 1992).

The specific emission is caiculated bydividing the sum of the basic volumes ofgas released in the volume of influenceby the number of tonnes of coal extracted -(or by the advance of the face). This ratio Jtherefore represents an estimate of thevolume of CH4 likely to be released pertonne of extracted coal (or per meter ofadvance).

- 2 -

Under the geological and operatingconditions of the Lorraine Coalfield, thismethod gives a satisfactory degree ofprecision. The discrepancy between thepredicted emission and the emissionactually observed is in the order of ± 10to 20% and only goes outside this rängein very special cases (see Figure 3).

However, it must be remembered that thiscaiculation to predict the specificemission only gives an average value forthe lifetime of the face, after its initialstarting period.

DYNAMIC PREDICTION USING ASTATISTICAL APPROACH

Observation of the firedamp balance in alarge number of faces has shown thepertinence of the hypothesis already made(British Coal, 1988) by which the gasemission over a week depends not onlyon the week's advance, but also on theadvance of a certam number of previousweeks.

In order to carry out investigations aimedat confirming this idea, a data base for thefiredamp balance of a representative setof coal mines in the Lorraine Basin hasbeen created. The faces selected met thefollowing conditions: dip of 15° to 30°,opening of 2.5 to 4.0 meters, caving,length of face 200 to 300 meters, advancerate 10 to 40m/week. In order todetermine the relationship between theemission for a given week (n) and theadvance of a previous week (n-x), linearregressions were carried out. The results,which are based on all the data in the lifeof the faces since their starting, arepresented in Figure 4. Since they did notprove conclusive, it was noted that onefactor in particular strongly influencedthe statistical relationship between thesevariables. This factor is the increase inthe firedamp emission after the face

begins, corresponding to the expansion ofthe volume of influence.

An analysis was then carried out of thecorrelation between the specific emissionand the cumulated advance of the face.This was used to determine the averagevalue of a "critical length" for theLorraine Basin faces i.e. the cumulatedadvance for which the face is in steadystate with regard to firedamp. The valuefound is about 220 m (see Figure 5).

Elimination of part of the datacorresponding to cumulated advances ofless man 220 meters considerablyimproves the results of the previous linearregressions (see Figure 6).

After carrying out a statistical Student-Fischer test to check that the correlationsare significant, it can be said that thevolume of firedamp released during theweek depends in practice on the advanceduring that week and those during theprevious two weeks. In general, thisdependence is much greater for theadvance of the week in question anddecreases with time, which is inaccordance with the physical firedampemission modeis (Günther, 1965;Ettinger, 1966; Airey, 1971).The emission DQ for week n can thus beexpressed äs follows :

Dn=^s[.anAn+:in-lAn-l+^n-2An-2+ca^

where:

— Dn is the emission for week n, in m^,— D§ is the specific CH4 emission of

the face, in m3 per meter of advance,— A^, AQ_^, Aß.2 are the advances of

week n and the two previous weeks,— a , a^.i, a^. , C^are constants.

Using a multiple regression analysisbetween the weekly advances An, A^.i,

- 3 -

Aii- 2 (which are independent variables)and the weekly emission D^ (which is thedependent variable), the values of theconstants in equation (l) were found for aset of 14 faces.

Significant correlations with a physicalmeaning (positive values of constants)were obtained for 7 faces. The results arepresented in Table l. It should be notedthat the majority of the faces for whichnon-significant correlations ornonsensical constants were obtained areeither faces affected by other workings orfaces for which very intensive drainagehas occurred.

Given the relatively low Variation in thevalues of the parameters obtained,average values were taken (see Table l).Thus, the following formula can be usedto predict the firedamp emissionaccording to the advance of the facebeing considered:

Dn=Ds[306An+150An-l+75An-2+5470)(m3) (2)

where:

Dn is the expected volume offiredamp for week n, in m^,D§ is the specific emission, obtainedusing the method describedpreviously (chapter 2), in m3 of CI-U.per meter of advance,AQ is the planned advance for week n,in m,AQ.I and A^-2 are the real advancesfor weeks n-1 and n-2 in meters.

A similar formula can be obtained byexpressing the emission for week n äs afünction of the tonnage produced duringweeks n, n-1 and n-2.

The use of this method for several faceshas produced very promising results, even

for faces affected by other workings orthose with intensive drainage. Figures 7and 8 give a comparison, by way ofexample, of the weekly emissionsactually observed for two faces and thoseobtained using the prediction method.

However, this method still needs furtherexamination. The values of theparameters obtained are average valuesfor all of the Lorraine Basin. It should bepossible to find the values of theseparameters for a specific area or mine,using the regionalized variables method,for example.

Finally, the values of the constantsobtained need to be related to physicaldata, such äs stratigraphic data.

PHYSICAL MODELLING PROPOSALFOR FIREDAMP EMISSION IN A

FACE.

It is quite obvious that, although thestatistical approach proposed above givesinteresting results, it only applies to theperiod after the face starting phase.

This limitation of the method has led tothe search for another model which willenable a dynamic prediction of thefiredamp emission to be made throughoutthe life of a face.

The model proposed below is based onexperience acquired in the field of gasemission. It is therefore a physical modelbased on experimental results.

If we observe the CH4 emission of a coalor rock bed located at a certain distancein the roof or the floor of the face äs afünction of the displacement of the face, acurve similar to that given in Figure 9 isobtained (CEC, 1980; Airuni, 1981). Theemission rate here is plotted against time

- 4 -

which, in the case of a regulär advance, isthe same thing.

The following can be noted :

• a delay in the beginning of emission,corresponding to extension of thevolume of influence up or down to thebed considered,

• a very rapid increase in intensity,corresponding to relaxation of thebed,

• then, after reaching a maximum, amuch slower decrease.

For a given period of time, the volume ofCH4 released by the bed will be thatrepresented by the shaded area inFigure 9.

This physical phenomenon can be easilymodelled by the following mathematicalexpression (for the roof):

for t < t ^ : ^(0=0

for t > to :

r R R i2i ^.-[Log(t-tQ ) -m J ^

^——————^_^ _______________J

R 22 (o )^2)t o (t- Iß )

the delay t^, the shorter the time required

to reach maximum intensity t^, and thehigher the maximum emission speed. Inother words, a bed in dose proximityreleases its gas earlier, faster and at ahigher flow rate. A distant bed releases itlater, for a longer period and with a lowerflow rate.

Figure 10 represents the emission curvesfor 3 beds located at different distances.

In order to represent these phenomena, it

can be considered that the values of t^

and t obey parabolic laws äs a functionof the distance between the bed i beingconsidered and the mined seam, ( (i):

t^a-^d)]2 (4)(parabolas l and l' on the Figure 10)

1^(1)= ß^d^i)]2 (5)(parabolas 2 and 2' on the Figure 10)

and that the maximum value of thefunction P1 obeys a hyperbolic law :

(3)

This is a log-normal distribution where :

— P is the gas release rate for that partof the bed being considered (roof) inm^/day or m^/week,

— t is the time elapsed after the face haspassed at the vertical of the pointbeing considered,

— t^, m11 and ö11 are parameters.

It can also be said that the closer the bedis to the seam being mined, the shorter

-[^[t^i)]-!!!1^)]

^Pl-—————————r R ,2——————— J=r K ^^2 [o (i)]

l(6)

1^(1)= ^ (D- tod) ]2

(characteristic of the log-normal distribution) (7)

where ö^, R and y11 are parameters.

In these last expressions, all the

parameters tg , t, , m and o are

- 5 -

Dexpressed äs a function of i, the indexrepresenting the different beds containedin the volume of influence. Theseexpressions are the same for the roof andfloor beds, but with different parametervalues (F Indexes for the latter).

We know that the face starting phasecorresponds to an expression of thevolume of influence while the subsequentphase corresponds to its translation.During the first phase, there are longerdelays before emission begins from thatpart of the bed being considered than inthe second phase. This phenomenon canbe modelled by parameterizing parabolasl, 2, l' and 2' in the following way (forthe roof):

(T <+£ (8)

where :

- 0.0 is a constant,- L is the cumulated advance of the

face from the starting,- Lc is the critical advance length,- e = 0. l or 0.05 for example.

The parabolas 3, 4, 3' and 4'corresponding to the starting period tendrespectively towards parabolas l, 2, l'and 2', which correspond to thesubsequent phase and are steady. TheParameters ß and 7 are taken to beconstant (ßp and 7o)-

Having mathematically expressed all thephysical phenomena, the total CH4emission during a given period, on agiven day, j, for example, canbe expressed äs follows :

100

R i+ SP„( i )£ A(k) J:k+ l f.^OdtJ-k lk = l

+fpF( i )£A(k) JI^f^Odt] (9)S PF (i) ii = l k = l

where:

- Dg is the specific emission caiculated(see chapter 2) in m^/m of advance,

- PMS is the share of this specificemission produced by the minedseam, in %,

- PR (i) is the share of this specificemission produced by bed number i atthe roof, in %, S

- Pp (i) is the share ^of this specificemission produced by bed number i atthe floor, in %,

- R is the number of beds at the roof,- F is the number of beds at the floor,- A(j) and A(k) are the face advances

on days j and k respectively, in m,

- ^(t) is given, for bed i, by expression

(3) and f^t), by a similar one for thefloor.

with:

- 0^(1), t^(i) and m^(i) obtained byresolving the System of equations (4)to(8),

- (^(i^t^andm^i) obtained byresolving a similar System for thefloor.

The CH4 emission, G(j) is expressed äs a

function of six parameters, a^,ß^,y for

the roof and o , ß , y^ for the floor.

The practical method consists inadjusting the parameter values on areference face by minimizing the

- 6 -

discrepancies between the caiculationsand measurements for the working. Theleast error squares method can be used forthis, for example. In practice, if the CI-L).emissions are caleulated on a day-to-daybasis, they can be marred by errors. Forthis reason, it is preferable initially toconsider a weekly CH4 emission.

After obtaining the Optimum values forthe six parameters and checking theirvalidity, the same formulae cansubsequently be used to make realpredictions for other faces.

CONCLUSIONS

Here, two methods are presented for adynamic prediction of firedamp emissionin longwalls :

• a statistical method has been definedand has already given interestingresults for several workings, butnormally cannot correctiy representthe starting period for the face, .-'

• the mathematical bases have been laidfor an empirical method closer to thephysical reality of the phenomenon.This method will need to be validatedduring implementation.

The emission of firedamp around amining structure is an eminently complexproblem. It is very difficult to equate theproblem äs a whole, given the number ofparameters involved and, in particular,the practical difficulty of apprehendingthe values of these parameters. Thus,what is proposed here is a pragmaticapproach to the problem.

These methods can provide the operatorwith a prediction of the firedampemission. He needs tools to optimize themining work, that is, to adjust the human

and matenal resources available to thereal production capacities of a working.

The prediction methods described hereare an improvement in this respect andfinally should result in better productivityfor workings äs well äs greater safety.

ACKNOWLEDGEMENTS

The work whose results are presented herereceived financial backing from the CoalDirectorate of the Commission of EuropeanComunities, from the French Ministry of Industryand from Charbonnages de France.

We would like to thank the engineers andtechnicians from the Lorraine Coalfields for theirefficient assistance in carrying out this work.

BIBLIOGRAPHY'v

AIREY E.M., 1971, "A theory of gas emission inmining operations". International Conference onSafety in Mines Research Institutes, Donetzk.

AIRUNI A.T., 1981, "Theorie et pratique de lalütte contre le grisou dans les mines profondes",Editions Nedra - Moscou.

BOROWSKI J., 1969, "Les relations entre laproduction et le degagement en taille", Travauxdu GIG, Publication n° 472, Katowice.

BRITISH COAL CORPORATION, 1988, "Short-term forecasting of methane emission levelsusing continuously monitored data", CECResearch Project 7220, AC/832.

BRUYET B., 1967, "Les variations de teneursdans les retours d'air de tailles - Influence desfacteurs d'exploitation", Publication CERCHARn° 1815 - Verneuil-en-Halatte.

CEC (Coal Directorate of the Commission of theEuropean Communities), 1980, "FiredampDrainage", Verlag Glückauf, Essen.

ETTINGER I.L., 1966, "Capacite d'adsorptiondes gaz par le charbon", Edition Nedra, Moscou.

GÜNTHER, 1965, "Etüde de la liaison gaz-charbon", Revue de l'Industrie Minerale, volume47, n° 10, Saint-Etienne.

- 7 -

INERIS, 1992, "Applications de la connaissancedu mode de degagement du grisou pour laconduite des exploitations rapides", Projet derecherche CCE 7220 - AC/319.

JEGER C., 1980, "Elements nouveaux dans laprevision du degagement de grisou dans lestailles. Journees d'informaüon : grisou, climat etaerage dans les charbonnages de laCommunaute Europeenne", Luxembourg.

KAFFANKE H., 1980, "Prevision ä moyen termedu degagement de grisou dans les voies deretour d'air des chantiers", Journeesd'information : grisou, climat et aerage dans lescharbonnages de la Communaute Europeenne -Luxembourg.

WINTER K., 1971, "Application de l'analysestatistique des methodes au caicul previsionnelde degagement de grisou". InternationalConference on Safety in Mines ResearchInstitutes, Donetzk.

Weekly ilredamp emissionU.E. Reumaux - Face Frieda 5 South block l 960/1036

900 —100.0

(^E§ 700 -

£•gä 500 -+- averaee emission predicted

"EB 300 •ü'S ;

l ' " lÜhlliiill l> s= «c ic =a^ sp

advanceauvaiicc-———-» „ • - ,„ - r; - i- ,^ .*» ".• - Ty" 6... ^ r'"'• • —**» ** production"' •— \ / ^

Weeks

Figure l

90.0

80.0

70.0

60.0

0

50.0 S §« •5u u

40.0 § .§^ 0•B >-

30.0 « B-

22'20.0 g g

^10.0

0.0

DETERMINATION OFAVERAGE MULTIPL E REGRESSION PARAMETERS BETWEENWEEKLY EMISSION AND THE FACE ADVANCE

REGRESSION ON DATA WHOSE CUMULATE D FACE ADVANCE IS GREATER THAN 220 M

Parameter valuesLongwall Mine Correlation

coefißcientr2

Degrees offreedom

Weekn Week n-1 Weekn-2

«n-2Louise 11 bis Reumaux 0.62 31 223 124 76

Irma II sud Reumaux 0.47 40 143 41 50

Louise I bis Reumaux 0.84 18 399 219 56

Kpan.2 Forbach 0.64 48 344 167 127

H2pan.l Forbach 0.93 10 413 268 66

Albert 9.6.0 La Houve 0.77 12 451 116 46

Frieda 5 Sudpan.1

Reumaux 0.50 36 168 116 112

TABLE1- 9 -

Model for predicting the average CH4 emission in a face

Degassing volume Intensity of degassing

Figure 2

-10-

Comparison between specific emissionsemissions predicted and actually observed

20 30 40 50 60 70

Specific emission actually observed Dr (m3/t)

Figure 3

Faces studie dLinear regression:

emission for week n = f (advance ofweek n-x)(all data)

n-2 n-3 n-4 n-5 n-6 n-7 n-8 n-9

Week to which regression applies

Figure 4

- l l -

-o—— Louisellbis

-^—— Irma II Sud

-<i—— Louiselbis

-ü—— Fried» S Sud p.1

-x—— Cpan.1

--3—— Kpan.2

-• --• Kpan.5

-+• •• H2pan.1

-• »• •• Alben 9.6.0.

-A- - - Albert 3.1.2.

-x--- Alben 9.5.1.

-• -•• Theodore 2.9.6.P.1

Variation of specific emission äs a functionof the cumulated advance

100.0 200.0 300.0 400.0 500.0 600.0

Cumulated advance offace L (m)

Figure 5

Faces studied

Louis« II bis

Irma II Sud

Louis« l bis

Frieda 5 Sud pan.2

Cpan.1

Kpan.5

H2 pan.1

Albert 9-5-1

Frieda 5 Sud pan.1

regressio n curv e

— • predictio n interva l

— - predictio n mterva l

Linear regression :emission for week n = f (advance ofweek n-x)

(data for which the cumulatedadvance is greater than 220 m)

n-1 n-2 n-3 n-4 n-5 n-6 n-7 n-8 n-9

Week to whic h regressio n applie s

Figure 6- 12-

Faces studie d

Louis« II bis

Irma II Sud

Louis« l bis

Frieda 5 Sud p. 1

Cpan.1

K pan.2

K pan.5

M 2 pan.1

Albert 9.6.0.

Albert 3.1.2.

Albert 9.5.1.

Theodor« 2.9.6.P.1

Variation of the CH4 emission rate äs a function of time

Figure 9

Shape of the emission curve äs a function of the vertical distanceof the beds from the mined seam.

Figure 10

- 13-

(l) and (2): maximum parabolas(period after the starting phase)

(3) and (4): parabolas duringthe starting phase.

Prediction ofweekly CH4 emission äs a function ofthe advanceFrieda 5 face South block 2

700

^ 600B

2 500^,

.1 400'5.2

§ 300-«fffiÜ 200>t3

| 100

0

emission measured prediction

47 49 51 89/1 3 5 7 9 11 1348 50 52 2 4 6 8 10 12

Weeks

Figure7

Prediction ofweekly CH4 emission äs a function ofthe advanceAlbert 9-5-1 face, district 03.

l-'U

«rE0

2 loo,

^g*BM

«

S 50

u21

S^

n

-• -emission measured —-prediction• ' •

• ', ' / \\

" ^. ' M^• -• 'Y ^/'^. l / ^^^ ' ^^ ;

\'~~~~^'".. /'' '', ' • . i' .•'.X ^ ; '. ~. ;'-''

*—»"' •' . K. '' . i• • : ~- /1 : 1 1 ! ; ' 1! . "-* ' ! ! . . ' • 1 , : ^^" 1; ' , '

'

. . , ; , . , , , ' , . , , . , ! , • - i . . ' . • . \ ; . , . , , , , ; .

35 37 39 41 43 45 47 49 51 89/1 3 5 7 9 1136 38 40 42 44 46 48 50 52 2 4 6 8 10

Weeks

Figure 8

-14 -